The Potential of Autonomous and Semi-Autonomous Vehicles in Supporting the Sustainable Development of Road Freight Transport

Abstract

Road freight transport (RFT) faces growing pressure from increasing freight demand, stricter environmental requirements, and persistent driver shortages. Automation technologies (ATes)—especially semi-autonomous driving—are increasingly viewed as a practical pathway toward improving the sustainability performance of freight operations; however, their effects depend strongly on infrastructure and operational conditions. This study evaluates the sustainability potential of autonomous and semi-autonomous trucks through an integrated framework combining (i) a structured review of technical and regulatory developments, (ii) surveys of transport enterprises (TEes) and road users (RUs), (iii) SWOT/TOWS analysis, and (iv) a cost minimization logistics model that links operational feasibility to infrastructure readiness (IR). The proposed model minimizes cost per tonne-kilometre and introduces an Infrastructure Readiness Score (IRS) to represent the share of a route that can be operated in automated mode; it also accounts for fuel savings from platooning and higher maintenance and capital costs of semi-autonomous vehicles (SAVs). Results indicate that, as IRS increases, semi-autonomous operations achieve higher daily mileage and lower unit costs, with a break-even point at approximately IRS ≈ 0.125. Beyond this threshold, unit costs decline from EUR 0.0433 to EUR 0.0348 per tonne-kilometre as IRS rises toward 0.6, after which further infrastructure improvements yield diminishing mileage gains. These cost and utilization improvements imply sustainability benefits via improved energy efficiency and reduced emissions intensity per tonne-kilometre. Nevertheless, survey evidence highlights major adoption barriers, including insufficient IR, regulatory uncertainty, technological reliability concerns, and limited public trust in fully autonomous systems. Overall, the findings support semi-autonomous trucking as the most feasible near-term stage of transition, while emphasizing that infrastructure upgrades and governance mechanisms are critical for scaling sustainability gains.

1. Introduction

Road Freight Transport (RFT) continues to play a central role in modern economies [1,2,3]. It ensures the uninterrupted movement of goods, supports industrial activity, and underpins increasingly complex supply chains [4,5,6]. Owing to its flexibility and operational accessibility, this mode of transport remains dominant in inland freight distribution [7,8,9]. At the same time, however, the sector faces a growing number of challenges [10,11,12]. Rising freight volumes, increasing energy consumption, greenhouse gas emissions, congestion, safety concerns, and a persistent shortage of professional drivers all contribute to mounting pressure on Transport Systems (TSs) [13,14,15]. These developments have intensified discussions on how RFT can evolve in line with sustainable development goals [16,17,18]. Until relatively recently, the idea of vehicles operating without direct human control was often regarded as unrealistic [19,20,21]. Yet the broader trajectory of technological progress suggests otherwise. Innovations once perceived as unattainable have repeatedly become part of everyday practice [22,23]. The historical evolution of transport technologies—from early mechanical solutions to highly digitalized systems—illustrates a continuous process of adaptation and transformation [24,25,26]. In this context, autonomous vehicles (AVs) and semi-autonomous vehicles (SAVs) may be seen as a natural step in the ongoing development of mobility and logistics systems rather than as a sudden technological disruption [27,28,29]. The increasing visibility of automation technologies (ATes) has shifted the focus of both academic research and industry debate [30,31,32]. The question is no longer whether such technologies will influence freight transport (FT), but how this influence will unfold and what consequences it may bring [33,34,35]. While automation is frequently associated with improvements in efficiency and safety, its broader implications for sustainability are far more complex and require systematic examination [36].

Road freight transport is a fundamental element of modern supply chains. Its importance stems primarily from its high flexibility and the ability to deliver directly to the end user. Research conducted to date indicates that road transport plays a significant role in ensuring the continuity of production processes, supporting extensive logistics networks, and operating national goods distribution systems [1,2,3]. Furthermore, it should be emphasized that growing challenges characterize the freight transport market. These include, in particular, the growing demand for transport services, high energy consumption, rising pollutant emissions, and labour shortages. These factors are increasingly negatively impacting the functioning and efficiency of freight transport systems [13,14,15].

From an operational perspective, research to date suggests that autonomous and semi-autonomous vehicles significantly impact freight transport. This is primarily due to a more stable driving style, reduced driver errors, and improved fuel efficiency [37,38,39]. Furthermore, both empirical analyses and simulation studies indicate that these technological solutions can translate into better utilization of available resources and reduced transport emissions. It should also be noted that the scale of the benefits achieved varies and is largely determined by numerous technical, organizational, and market factors [40,41,42].

The main contribution of this paper covers three key areas. First, our own research conducted a comprehensive assessment of autonomous and semi-autonomous road transport, focusing on technological, organizational, and economic factors. Second, we proposed an Infrastructure Readiness Index (IRS) that indicates the feasibility of implementing automated driving, accounting for route-specific conditions. Therefore, this indicator allows for a more precise assessment of implementation potential. Third, the study integrates empirical research results with a strategic analysis (SWOT/TOWS) and a cost minimization model. This approach bridges the gap between theoretical considerations and the business practice of freight transport companies. In this way, we can answer the question of how automation can impact the operational efficiency and sustainable development of freight transport.

1.1. Background and Problem Statement

The transport sector remains one of the major contributors to global CO₂ emissions, with RFT accounting for a considerable share of total energy use and environmental impact [43,44,45]. The steady growth of freight demand—driven by globalization, expanding trade flows, and the rapid expansion of e-commerce—further intensifies environmental and operational pressures [46,47,48]. At the same time, transport enterprises (TEes) operate under conditions shaped by rising operating costs, labour shortages, and increasingly stringent environmental regulations [49,50,51]. Conventional strategies aimed at improving FT performance, including fleet modernization, route optimization, and organizational adjustments, continue to play an important role [52,53,54]. However, their ability to deliver long-term sustainability improvements may be limited. As a result, attention has increasingly turned toward technological innovations capable of generating more systemic change [55,56,57]. Among these, vehicle automation is widely regarded as one of the most promising developments [58,59,60]. From an operational perspective, AV and SAV may influence FT in several ways [37,38,39]. These include more stable driving patterns, a reduction in human error, and the potential for improved fuel efficiency and resource utilization [40,41,42]. Nevertheless, the magnitude—and even the direction—of these effects cannot be assumed a priori [61,62,63]. Whether such benefits materialize in practice depends on multiple factors, including technological maturity, IR, regulatory conditions, and economic feasibility [64,65,66]. In addition to technological considerations, ATes introduce a range of legal and institutional challenges [67,68,69]. Issues related to liability, safety standards, regulatory adaptation, and risk allocation require careful analysis [70,71,72]. From a technical standpoint, the effective functioning of automated vehicles relies on advanced system architectures incorporating sensors, radar technologies, cameras, and intelligent control mechanisms that together enable perception and decision-making processes [73,74,75].

1.2. Research Gap

Although academic interest in AVs has grown rapidly, much of the existing literature focuses primarily on passenger transport, engineering aspects, or isolated safety analyses. Comparatively fewer studies address RFT from an integrated sustainability perspective. Moreover, many prior contributions emphasize conceptual discussions or projected benefits rather than combining empirical evidence with modelling approaches. There remains limited understanding of how different levels of vehicle automation influence sustainability indicators in freight TSs, particularly when environmental, economic, and operational dimensions are analyzed simultaneously [76]. The transitional role of SAVs—which are more likely to be implemented in the near term—also requires further investigation. These observations highlight the need for research frameworks that integrate sustainability assessment, logistics modelling, empirical investigation, and strategic evaluation tools.

Published research on autonomous freight transport typically focuses on a single, selected area. Some focus on technological and regulatory issues, providing only conceptual or overview considerations for the development of automation systems. Other scientific publications present the results of surveys aimed at identifying stakeholder opinions, determining the level of social acceptance, and identifying barriers to implementing new solutions. Furthermore, another section of the publication addresses economic and operational efficiency. Specifically, these analyses focus on investment profitability, route optimization, and the overall efficiency of the transport system.

The research conducted by the authors of this article represents a comprehensive approach to the presented issue, coherently combining several complementary analytical perspectives. The research methodology is based on a structured literature review and surveys conducted among transport companies and road users. Furthermore, a strategic analysis was conducted using the SWOT/TOWS method, and a proprietary cost minimization model was developed that incorporates the Infrastructure Readiness Index (IRS). This combination of methods enables a comprehensive assessment of the problem under investigation, including technological feasibility, organizational preparation, and economic efficiency.

1.3. Aim and Structure of the Study

The aim of this study is to assess the potential of AVs and SAVs to support the sustainable development of RFT, with a particular focus on how different automation levels may affect operational efficiency and selected sustainability outcomes. In addition, the study examines the readiness of TEes to adopt ATes, acknowledging that implementation depends not only on technical feasibility but also on organizational capacity, perceived risks, and the broader institutional environment.

To address these objectives, a multidimensional approach is applied. The study combines a review of technological developments and selected regulatory considerations with empirical surveys conducted among TEes and road users (RUs), complemented by SWOT and TOWS analyses. Moreover, a refined logistics cost model is introduced to quantify how automation-related changes—together with IR conditions—may translate into differences in cost performance and resource utilization, which are closely linked to sustainability in FT. The remainder of the paper is organized as follows. Section 2 presents the theoretical and conceptual background. Section 3 describes the empirical design, analytical tools, and modelling approach. Section 4 reports the comparative results and model-based findings. Finally, Section 5 summarizes the main conclusions and outlines directions for further research.

1.4. Contribution to Research

The study conducted by the authors highlights several important elements that add value to the discussion on autonomous freight transport.

First, it proposes a coherent research approach that combines methods ranging from empirical research and SWOT/TOWS analysis to statistical analysis and cost modelling. This approach allows for a simultaneous focus on technological, organizational, and economic aspects, which, as previous studies have shown, have been considered separately.

Second, the authors introduce a cost-based analytical model that incorporates the Infrastructure Readiness Index (IRS). This allows for a more concrete demonstration of how infrastructure condition impacts the efficiency of semi-autonomous transport.

Third, the study also establishes a cost-effectiveness threshold for this type of solution (IRS of approximately 0.125). It should also be emphasized that this approach allows for the identification of practical aspects, which are often lacking in more theoretical analyses that dominate the literature. Fourth, this article presents the results of empirical research conducted among transport companies and road users, thereby increasing the research’s credibility. Furthermore, it should be noted that the use of statistical tools such as Pearson and Spearman correlations, analysis of variance (ANOVA), Student’s t-test, and Kruskal–Wallis test allowed not only to describe the phenomena but also to identify statistically significant relationships and differences between the study groups.

Fifth, the study addresses the problem from two complementary perspectives: industry and societal. This means that, in addition to issues related to the organization’s preparation for implementing semi-autonomous solutions, the level of their acceptance among road users was also considered. This combination is significant because in previous studies, these two dimensions were most often analyzed separately.

In summary, the paper contributes to a better understanding of semi-autonomous transport as a transitional solution, not a final one. The authors particularly emphasize that the widespread adoption of semi-autonomous vehicle technology does not depend solely on the technology itself. Based on the research results, it can be concluded that factors such as the state of infrastructure, the regulatory framework, risk perception, and the level of user trust are also very important.

2. Theoretical and Conceptual Background

2.1. Application of Autonomous and Semi-Autonomous Vehicles

Continuous technological development has a profound impact on the quality of human life [77,78,79]. The emergence of the first motor vehicles marked a significant breakthrough, overcoming major barriers related to the transportation of goods and the movement of people over long distances [80,81,82]. This represented an important milestone not only for technological progress but also for socio-economic development [83,84,85]. Driven by ongoing innovation and the persistent pursuit of improvement, the twenty-first century witnessed the gradual introduction of AVs and SAVs on urban roads and highways [86,87,88]. The application of such technologies in TSs offers a range of potential solutions that may positively influence traffic safety and contribute to the reduction in accident-related casualties [89,90,91]. Among the anticipated benefits frequently highlighted in the literature are lower CO₂ emissions, reduced delivery times, mitigation of urban congestion, and improvements in overall transport efficiency [92,93,94].

At the same time, it should be acknowledged that technological progress is often accompanied by new challenges [95,96,97]. The introduction of AVs involves substantial costs associated with infrastructure modernization, as well as risks related to potential system failures, traffic disruptions, and labour market implications, particularly for professional drivers [98,99,100]. Although many of these challenges may appear manageable, their mitigation requires careful evaluation and systematic analysis [101,102,103]. To support the structured development and implementation of vehicle ATes, the National Highway Traffic Safety Administration (NHTSA) proposed a classification framework describing different levels of vehicle autonomy [104,105,106]:

• Level 0 (No Automation). At this level, the driver maintains full control over all vehicle functions, including acceleration, braking, and steering. The human operator is solely responsible for the safe operation of the vehicle. Level 0 represents the complete absence of vehicle automation. At this stage, no Artificial Intelligence (AI) or automated driving systems (ADSs) are involved in vehicle operation. Driving performance depends entirely on the skills, experience, and decisions of the human driver (HD), who is also fully responsible for any errors or incidents that may occur.
• Level 1 (Function-Specific Automation). This level includes automation systems that support one or more specific driving functions operating independently of one another. The driver retains full control over the vehicle and remains solely responsible for its safe operation. Automated systems may intervene only under predefined conditions, either assisting the driver during normal driving situations or supporting vehicle control in potentially hazardous circumstances. Typical examples of Level 1 technologies include braking assistance systems, Electronic Stability Control (ESP), and Adaptive Cruise Control (ACC). Although these systems enhance driving comfort and safety, overall vehicle control continues to rest with the HD. Nevertheless, Level 1 automation already introduces elements of active (functional) safety into vehicle operation.
• Level 2 (Combined Function Automation). This level involves the automation of at least two primary vehicle control functions designed to reduce the driver’s workload. While certain driving tasks may be delegated to automated systems, the driver remains responsible for supervising the driving environment. Continuous monitoring of the surroundings is required, and the driver must be prepared to immediately resume control whenever necessary. Examples of Level 2 systems include ACC integrated with lane-keeping assistance technologies. These solutions provide partial driving automation but do not eliminate the need for active driver engagement. This level may be considered a continuation of the previous stages of automation, as the presence of an HD remains essential. Levels 0 to 2 collectively ensure that the driver retains overall control of the vehicle. These systems significantly support driving tasks and contribute to reducing the risk of collisions and road accidents by continuously assisting and monitoring selected vehicle functions. Technologies corresponding to these automation levels are already widely implemented in contemporary vehicles. Their adoption has been facilitated by the fact that they generally do not require extensive modifications to existing road infrastructure. At the same time, they offer noticeable improvements in driving comfort and road safety. Overall, the degree of automation at these levels is commonly regarded as providing more advantages than disadvantages, particularly in terms of operational practicality and safety enhancement.
• Level 3 (Conditional Automation). At this level, ADSs are capable of assuming full control over key vehicle functions under specific traffic or environmental conditions. These functions typically include acceleration, braking, vehicle dynamics control, monitoring of the surrounding environment, and aspects related to operational safety. Although the system manages driving tasks, the HD is still required to remain available and prepared to take over control when requested, usually within a defined transition period. Compared with lower levels of automation, the role of the driver becomes more limited. The vehicle operates autonomously within certain scenarios, while human intervention occurs only when necessary. The reliable functioning of Level 3 vehicles depends on the integration of multiple advanced technologies, including sensor systems, cameras, and real-time data processing capabilities. In addition, effective operation requires compatibility with highly developed infrastructure and telematics systems that continuously exchange information with the vehicle. Vehicles equipped with Level 3 technologies are expected to contribute to improved road safety by reducing the likelihood of accidents associated with human error. Furthermore, such systems may expand mobility opportunities for individuals with physical limitations. From an economic and environmental perspective, Level 3 automation may also generate benefits. Through cooperation with telematics and traffic management systems, these vehicles can support smoother traffic flow, potentially reducing congestion, shortening delivery times, and lowering CO₂ emissions. It is also important to note that the proper functioning of Level 3 automation relies on continuous, high-speed data transmission. In this context, advanced communication technologies such as 5G play a critical enabling role.
• Level 4 (High Automation/Full Autonomy). Vehicles classified at this level are designed to continuously monitor road and traffic conditions and to perform all driving-related tasks autonomously throughout the entire journey. The system maintains full control over vehicle operation, including navigation, acceleration, braking, and safety functions, without requiring human intervention. At this stage, the role of the human occupant is limited to defining the destination. The driver is not expected to supervise the driving process and is generally unable to assume manual control during vehicle operation. ADSs function independently, regardless of whether the vehicle is transporting passengers or operating without human presence. Level 4 automation represents a stage of full vehicle autonomy, where direct human involvement in driving tasks is no longer required. The role of the human occupant is limited primarily to entering trip-related data or selecting a destination. In such a scenario, the HD is effectively excluded from vehicle operation. While this level of automation offers numerous potential advantages, its implementation is associated with significant challenges. In particular, the deployment of Level 4 vehicles would require extensive modernization and reconstruction of existing transport infrastructure. Moreover, concerns arise regarding potential labour market disruptions, especially in relation to professional drivers whose roles may become substantially reduced. Infrastructure adaptation costs are frequently cited as one of the major barriers. According to available estimates, the cost of installing sensor systems necessary to support automated driving environments may reach approximately USD 5000 per 100 m of roadway [107]. One of the aspects requiring particular attention concerns the issue of full trust in automated systems and AI. Questions related to system reliability, including the potential consequences of communication interruptions or failures in environmental monitoring systems, remain critical considerations in the discussion on fully AVs. Among the most frequently emphasized advantages of full vehicle automation is the potential for improved logistics performance. AVs may contribute to shorter delivery times and a significant reduction in carbon dioxide emissions. These benefits are often attributed to smoother driving patterns, which can lead to lower fuel consumption and more efficient energy use.

The scientific literature on autonomous and semi-autonomous transport can be divided into three basic categories. The first focuses on technological developments, including sensor systems, vehicle control architectures, and communication technologies. The second group of literature analyzes regulations and legal provisions, primarily concerning safety standards, scope of responsibility, and certification procedures. The third strand discusses organizational and economic aspects, focusing in particular on the impact of automation on the efficiency of the freight transport process, the cost structure, and environmental benefits. However, it should be noted that existing publications lack work that integrates these aspects. Consequently, there is a research gap regarding the interplay between technological conditions, infrastructure development, organizational factors, and the economic dimension on the effectiveness of implementing autonomous and semi-autonomous freight transport systems.

2.2. Technological Characteristics of Autonomous and Semi-Autonomous Vehicles

Having outlined the different levels of vehicle automation, it becomes evident that the proper functioning of autonomous systems relies on the interaction of multiple technological components [108,109]. These include advanced control systems, sensors, cameras, and radar technologies that continuously cooperate through the rapid transmission and processing of data obtained from the surrounding environment and, in some cases, from other vehicles [110]. SAVs, typically associated with automation Levels 0 to 2, where the driver remains responsible for decision-making, are also equipped with a range of safety-enhancing systems [111]. These technologies are designed to support vehicle control and improve road safety without eliminating the role of the HD [112,113]. One of the earliest and most significant safety systems introduced in modern vehicles was the Anti-lock Braking System (ABS) [114,115]. This system prevents wheel lock during sudden braking, thereby allowing the driver to maintain steering control. As a result, ABS contributes to improved vehicle stability, reduced braking distances under certain conditions, and enhanced tyre adhesion to the road surface [116]. Importantly, the system is activated only during braking events initiated by the driver [117,118]. The ABS technology was first implemented in 1978 in a Mercedes-Benz passenger vehicle. Since July 2004, the installation of ABS has become mandatory for all newly manufactured passenger cars within the European Union [119,120]. Additional systems commonly implemented in SAVs include Acceleration Slip Regulation (ASR) [121,122,123] and ACC [124,125,126]. The ASR system is designed to prevent wheel slip during vehicle start-up, particularly on low-adhesion surfaces.

The system activates automatically when a loss of traction is detected, reducing engine torque and adjusting power delivery to the wheels in order to maintain vehicle stability [127]. ACC represents another important driver assistance technology [128]. This system enables the vehicle to maintain a predefined speed while simultaneously regulating a safe distance from the preceding vehicle [129]. Its operation is typically supported by radar sensors installed in the vehicle structure [130]. SAVs are also frequently equipped with lane-keeping assistance systems and automated lighting technologies [131].

These solutions adjust vehicle behaviour to prevailing traffic and environmental conditions [132]. Their functioning is generally based on camera systems, most often positioned near the windshield, which continuously monitor road markings, surrounding vehicles, and lighting conditions [133]. The reliable operation of fully AVs requires a significantly more complex technological configuration. Such vehicles depend on the integration of multiple sensing and positioning technologies, including cameras, radar, lidar (laser-based sensors), odometric sensors, ultrasonic sensors, GPS modules, and high-speed communication networks such as 5G [134,135,136]. These components are coordinated by high-performance onboard computing systems responsible for real-time data processing and decision-making. In addition to vehicle-based technologies, the deployment of higher levels of automation necessitates advanced road infrastructure equipped with intelligent sensing and communication systems capable of exchanging data with vehicles [137]. The elements described above are essential for the proper functioning of vehicles operating at automation Levels 3 and 4. A more detailed description of the technological structure of AVs is provided below [138].

An AV may externally resemble a conventional passenger car or a heavy-duty tractor unit. The vehicle’s autonomous capabilities are enabled by an array of sensing technologies, among which 360-degree camera systems play a particularly important role [139]. These cameras continuously monitor the vehicle’s surroundings, providing visual data necessary for environmental perception [140]. Camera-based systems are highly effective in detecting traffic lights, road signs, and lane markings. However, their performance may be affected by adverse or rapidly changing weather conditions, such as fog, rain, or snow [141]. In addition, object recognition algorithms may encounter difficulties in situations characterized by low contrast or when the colours of objects are similar to the background. Under such conditions, the reliability of visual detection may decrease, potentially increasing the risk of misinterpretation of the driving environment [107]. Another essential component of AV technology is the use of radar and lidar systems [142,143,144]. Radar technology itself is not a recent innovation, as it has long been applied in fields such as aviation and maritime navigation. The operating principle of radar systems is relatively straightforward. Radar sensors emit high-frequency electromagnetic waves (radio waves), which are reflected by surrounding objects. By measuring the time interval between signal transmission and reception, the system is able to determine the distance and relative position of detected obstacles. Radar systems offer several important advantages, including long detection ranges—typically up to approximately 200 m—and broad spatial coverage provided by the emitted wave beams. Furthermore, radar performance is generally less affected by weather conditions, and different radar systems can operate simultaneously with minimal mutual interference [141]. Radar systems may initially appear to represent an ideal sensing solution, as they are capable of detecting obstacles and, when integrated with appropriate control software, can support automated responses such as emergency braking [145]. However, radar technology also exhibits certain limitations. In particular, the effectiveness of radar signals depends on the reflective properties of detected surfaces, with stronger reflections typically obtained from metallic objects. For this reason, AV architectures commonly incorporate lidar systems as a complementary sensing technology. Lidar, often described as a laser-based detection system, operates on a principle similar to radar. Instead of radio waves, however, lidar uses significantly shorter light waves emitted by laser sources [146]. This enables more detailed environmental mapping and higher-resolution object detection under many operating conditions [147]. In lidar systems, the light source is a laser that emits pulses directed toward surrounding objects. The reflected light is subsequently captured by photodetectors, while a microprocessor-based unit analyses both the returned signal and the time required for its reflection. This process enables the precise determination of object distance and spatial positioning. To generate a three-dimensional representation of the environment, lidar systems typically employ multiple synchronized laser emitters operating in coordination with rapidly moving or rotating mirrors. Such configurations allow the sensor to scan the surroundings and construct detailed 3D maps of the vehicle’s operating environment. Despite their advantages, lidar systems also present certain limitations. One of the primary constraints is their relatively shorter detection range, typically estimated at approximately 100–150 m. In addition, the performance of lidar sensors may be affected by light reflection phenomena and challenging environmental conditions. Another consideration is the comparatively high cost of such systems, with prices often cited at around USD 500 per sensor unit. Ultrasonic sensing systems also play a significant role in the operation of AVs. The working principle of these systems is conceptually similar to that of radar technologies. However, ultrasonic sensors operate using high-frequency sound waves, generally above 20 kHz. A key limitation of ultrasonic systems is their short operational range, usually not exceeding approximately 10 m, which restricts their application primarily to low-speed maneuvers [148]. For this reason, ultrasonic sensors are most commonly used in parking assistance and short-range obstacle detection scenarios [149].

All sensors and detection systems installed in AVs must operate in close coordination to ensure reliable performance [150,151,152]. To enable such integration, AVs are equipped with high-performance onboard computing units responsible for analyzing and processing the large volumes of data generated by the sensing systems [153,154,155]. It is estimated that the sensors of an AV may produce between 1 and 19 terabytes of data per hour [156,157]. This highlights the scale of information processing required for real-time environmental perception, decision-making, and vehicle control [158,159,160]. The acquisition, storage, management, and labelling of such large volumes of data therefore require carefully designed computing infrastructure [161,162,163]. This infrastructure must account for the demands of cloud-based processing systems as well as traffic management and coordination centres responsible for supporting real-time data exchange and analysis [107].

The scale of data processing required for the proper operation of AVs is substantial [164,165,166]. In this context, it is also important to consider the role of transport infrastructure and the sensing technologies embedded within it [167,168,169]. Although infrastructure components do not form part of the vehicle itself, their presence is often essential for ensuring reliable and efficient system performance [170,171,172]. Each of the sensing and detection systems described above is subject to certain limitations and potential error margins [173,174,175]. To mitigate these constraints, the exchange of data between vehicles and their surrounding environment has become a key area of development [176,177,178]. In particular, Vehicle-to-Vehicle (V2V) and Vehicle-to-Infrastructure/Infrastructure-to-Vehicle (V2I/I2V) communication systems are designed to enhance situational awareness and reduce the likelihood of perception or decision-making errors [179,180,181]. Through continuous data sharing, these communication frameworks can significantly improve the reliability and safety of ADSs [182,183,184].

Communication systems developed for AVs should also be compatible with conventional, non-AVs [185,186,187]. Such interoperability can further enhance the ability to anticipate, interpret, and model the behaviour of different RUs operating within mixed-traffic environments [188,189,190].

Effective cooperation between vehicles and transport infrastructure plays a critical role [191,192,193]. The reliable functioning of these systems depends on advanced telematics solutions supported by a network of sensors and detection technologies, including cameras and radar systems [194,195,196].

2.3. Deployment and Early Applications of Autonomous Vehicles

Vehicles corresponding to automation Levels 0 to 2 are already widely present on public roads. The majority of newly manufactured vehicles are equipped with driver assistance technologies such as lane-keeping systems, blind-spot monitoring, and ACC [197,198]. In addition, premium vehicle segments increasingly offer automated parking functions and other advanced support systems. The implementation of these technologies has not created significant operational challenges. This is largely due to the continued presence of the HD, who remains responsible for supervising vehicle operation and is able to assume full control at any moment. Furthermore, in the event of a collision or traffic incident, legal responsibility continues to rest with the driver [199,200,201].

More complex challenges arise with the introduction of higher levels of automation, particularly Levels 3 and 4. At Level 3, the driver may still intervene under certain conditions, although their role becomes more limited. Level 4 automation, by contrast, removes the possibility of direct driver control entirely. The inability of a human operator to resume full control represents only one of several issues associated with these automation stages. Numerous barriers remain, primarily related to IR, interaction with other RUs, and the adaptation of legal and regulatory frameworks.

From a regulatory perspective, one of the most critical challenges concerns vehicle approval for road use and the determination of liability in the event of accidents. In situations where a vehicle is operated by an HD, responsibility for a collision typically rests with the driver, provided that the incident was not caused by a vehicle defect but, for example, by an unintentional violation of traffic safety rules [202]. This is because the driver initiates vehicle movement and maintains control over the course of the journey. In the case of fully AVs, however, situations may arise in which a human operator is not physically present in the vehicle involved in a collision. Under such conditions, the question of liability becomes considerably more complex, particularly with respect to determining who should bear responsibility for any resulting damages. This issue has been addressed through a relatively straightforward principle. In the event of an accident, it is necessary to determine which entity exercised control over the vehicle at the time of the incident. If it is established that an HD assumed control of the vehicle—despite the absence of a requirement to intervene—responsibility for the consequences rests with the driver. Conversely, if the vehicle was operating in automated mode and the driver did not interfere with the driving process, liability may be transferred to the vehicle manufacturer or the software developer, depending on the identified cause of the accident.

Insurance for AVs represents a challenge comparable to the determination of liability in the event of accidents. Insurance providers offering coverage for automated vehicles often require comprehensive access to data recorded by onboard computing systems. Such data may be essential for reconstructing the circumstances of a collision, including whether the vehicle was operating in automated mode and whether the driver intervened when prompted by the system. Another important consideration relates to the need for professional adaptation within the insurance sector. The introduction of AVs may necessitate the retraining of personnel, particularly loss adjusters and technical experts responsible for accident assessment. The widespread adoption of AV technologies has the potential to significantly reshape the insurance market. In the long term, insurance premiums may decrease if automated vehicles, once technologically mature and reliable, contribute to a reduction in the frequency of collisions and traffic incidents.

Despite the challenges outlined above, research, development, and testing activities related to AVs continue to progress. Major automotive manufacturers, including BMW, Audi, Nissan, Tesla, and Mercedes-Benz, remain actively engaged in the advancement and evaluation of automated driving technologies. Interest in autonomous mobility is not limited to the automotive industry. Technology-oriented companies, most notably Google, have also played a significant role in the development and testing of AV systems. According to published reports, in August 2016 alone, Google’s autonomous test vehicles collectively travelled approximately 170,000 miles, of which around 126,000 miles were completed in fully autonomous mode. According to data reported by Google, approximately 94% of road accidents are attributed to human error. This observation is frequently cited as one of the key arguments supporting the future deployment of AVs, which are expected to contribute to a substantial reduction in accident rates through the minimization of human-related driving mistakes.

AV technologies are not limited to passenger cars. Particular attention should also be given to the heavy-duty vehicle segment, where the introduction of automation may generate substantial operational and economic benefits. In FT, automation is often associated with increased efficiency, reduced operating costs, and improved utilization of transport resources. Several leading manufacturers of commercial vehicles, including Daimler and Mercedes-Benz, are actively engaged in the research and development of autonomous truck technologies. During automotive industry exhibitions in 2019, Daimler announced its intention to introduce Level 4 automation within the following decade. The company also declared plans to invest approximately EUR 500 million and to create around 200 new jobs to support the development and commercialization of highly automated trucks.

Although the development of AV technologies is often presented in an optimistic light, it is important to acknowledge that their deployment is not free from risks. Notably, incidents involving AVs have already been recorded, including the first reported fatal accident. In this case, the automated system failed to detect a pedestrian crossing the roadway, resulting in a collision. Such events highlight the technological limitations that may still exist and underscore the need for continued research, testing, and regulatory refinement.

Another challenge that deserves attention is international transport within the European Union. It should be emphasized that, while the EU supports the free movement of goods, regulations governing the use of autonomous and semi-autonomous vehicles still differ across member states. Fundamental differences include vehicle licencing rules, liability in the event of accidents, insurance requirements, and data access. As a result, these discrepancies significantly hinder the smooth operation of international freight transport.

Therefore, it is essential to determine which regulations will apply when an automated vehicle crosses the border of a specific country. This will be particularly important in the event of a breakdown or accident, where liability may lie with various entities, such as the operator, vehicle manufacturer, or software provider. The lack of unified procedures and scope of liability will undoubtedly limit the use of such solutions by transport companies.

Therefore, to ensure the effective implementation of automated freight transport in international markets, it is necessary to develop common EU-level legal provisions to regulate certification, data sharing, and the scope of liability. Implementing semi-autonomous vehicles in the freight transport market requires attention to organizational changes and digital transformation within the enterprise. First and foremost, it is necessary to integrate advanced data systems, real-time monitoring tools, and decision-support systems. Transport companies must also systematically develop new skills in data management, system monitoring, and combining various technologies into coherent solutions.

It is also important to emphasize that this digital transformation can be both an opportunity and a challenge. Large companies with significant financial capital and technological resources tend to implement modern technology more quickly. Smaller companies, on the other hand, often face constraints stemming from limited investment capital, a lack of appropriate employee skills, or insufficient organizational flexibility. Furthermore, the transition to data-driven solutions is transforming the way companies operate. New business models are emerging, and logistics platforms and integrated supply chain management systems are playing an increasingly important role.

Therefore, the successful implementation of semi-autonomous transport systems depends not only on technological and regulatory conditions. In particular, attention should be paid to the company’s ability to adapt to the new, more digital reality, grounded in data analysis, modern systems, and constant change.

2.4. Integration of Autonomous Vehicles into Logistics Operations

Every autonomous car can be classified as a vehicle; however, not every AV is necessarily a car [203,204,205]. While the application of autonomous cars in transport is often discussed in the context of future developments and potential solutions to current mobility challenges, AVs have already found practical use in logistics operations [206,207,208]. One of the most common examples is the deployment of Automated Guided Vehicles (AGVs), widely used in production facilities, warehouses, and distribution centres [209,210,211]. These vehicles are typically employed for internal material handling tasks, such as transporting components or goods within controlled environments [212,213,214]. The adoption of AGV systems is frequently driven by the need to reduce operational costs, particularly those associated with labour-intensive processes [215,216,217].

AGVs operate differently from road-based AVs. In most cases, they follow predefined and fixed routes, which significantly simplifies navigation and control requirements [218,219]. As a result, their implementation generally involves lower costs compared with fully autonomous road vehicles [220,221,222]. Despite operating along fixed paths, AGVs are equipped with cameras and sensors that continuously monitor their surroundings and detect potential obstacles [223,224,225]. Vehicle control is supported by additional sensing and communication technologies, commonly including Bluetooth and wireless local area network (WLAN) systems [226,227]. Given their operation within restricted and structured environments, the limited maneuverability of AGVs is typically sufficient for achieving reliable and efficient performance [228,229,230].

AVs are currently more frequently applied—and are expected to remain more widely utilized—in logistics environments than fully autonomous cars in public road transport [231,232,233]. This tendency is largely associated with the regulatory complexities involved in approving AVs for operation on public roads [234,235,236]. In controlled settings, such as industrial facilities or logistics centres, vehicle operation is governed by a distinct set of internal rules and safety protocols [237,238,239]. Under these conditions, challenges related to insurance, liability, and legal responsibility are generally less pronounced [240,241,242]. By contrast, the deployment of AVs in open traffic environments requires the introduction and harmonization of extensive regulatory frameworks [243,244,245]. A wide range of legal provisions, operational standards, and compliance requirements must be defined and implemented. In conventional human-driven vehicles, the allocation of responsibility is relatively straightforward, as liability in the event of a collision or accident is typically attributed to the driver [246,247,248]. Greater complexity arises in situations where no human operator is directly responsible for vehicle control [249,250,251]. Consequently, the emergence of AVs has necessitated the development of dedicated legal and regulatory structures designed to clearly define and allocate responsibility within automated driving contexts [252,253,254].

One of the commonly discussed regulatory assumptions is that, during the initial stages of deployment, AVs may be subject to operational limitations [255,256,257]. These restrictions may include maximum speed thresholds, for example, limiting vehicle movement to approximately 60 km/h, as well as confining operation to road environments where interactions with vulnerable RUs, such as pedestrians and cyclists, are minimized [258,259]. Another frequently proposed requirement concerns the installation of event data recorders, often referred to as “black boxes” [260,261,262]. These devices are intended to continuously record and store vehicle operation data, which may be used for incident analysis and liability assessment [263,264,265]. In situations where a vehicle operates in autonomous mode while an HD remains seated behind the steering wheel, responsibility for vehicle supervision and potential consequences, including collisions or accidents, typically continues to rest with the driver [266,267,268].

AVs are still largely in the stages of technological development and real-world testing [269,270,271]. As a consequence, detailed legal regulations and comprehensive legislative frameworks governing their operation remain limited or are still evolving [272,273,274]. With further technological advancement and broader deployment, more specific regulatory provisions are expected to emerge [275,276,277]. These frameworks will play a critical role in ensuring legal clarity, operational safety, and adequate protection for all participants within the transport environment [278].

2.5. Data-Driven and AI-Based Approaches in Autonomous Freight Transport

Currently, there is a growing development of autonomous and semi-autonomous freight transport systems that utilize AI techniques. Initially, the focus was primarily on hardware aspects, such as sensors and control systems. However, software solutions are increasingly important. Machine learning methods, big data analysis, and advanced decision-making mechanisms are particularly important in this regard. This approach contributes to increased operational efficiency, improved safety, and enhanced reliability of the analyzed systems [279,280].

Currently, AI is used in many aspects of truck operation. Possible applications include fuel consumption forecasting, emission estimation, maintenance, fault diagnosis, and the development of autonomous driver assistance systems. Drawing on large sets of operational data from both vehicles and the transport infrastructure, AI-based models enable more precise and faster real-time process optimization. This ultimately reduces operating costs and the negative environmental impact of transport [281,282].

One of the most important elements of autonomous systems is the integration of data from various sources (sensors). Information from cameras, radars, lidars, GPS, and on-board diagnostic systems is collected and processed into a unified structure, providing a more complete picture of the surroundings and enabling better obstacle detection and appropriate action in challenging conditions. Research shows that such data integration, along with modern traffic planning algorithms, can significantly reduce collisions and increase the reliability of autonomous trucks, especially in more complex driving situations [283,284].

Another very important and rapidly developing area is predictive maintenance and vehicle health management. It should be emphasized that autonomous and semi-autonomous trucks generate large amounts of data, the analysis of which allows for the prediction of potential failures and improved maintenance planning. The use of machine learning models enables the estimation of component operating life (RUL), which also allows for proper maintenance planning. Consequently, this helps reduce the number of unplanned downtimes and financial losses. This approach is particularly important in long-haul transport, where vehicle reliability and availability directly impact the efficiency of the entire logistics system [285,286].

In terms of safety, the use of machine learning for risk assessment and the prediction of dangerous events is becoming increasingly important. Data-driven models are a helpful tool in detecting risky driving conditions, assessing rollover risks, and supporting more flexible vehicle control. For example, recurrent neural network models are used to predict the stability of trucks in motion, allowing for earlier responses and improved safety, for example, by adapting driving style or route. Therefore, AI-based safety systems can detect hidden patterns and risk factors that are difficult to identify using traditional methods [287,288].

Furthermore, the increasing use of autonomous trucks requires the development of more advanced methods for testing and monitoring their performance. Due to the complexity of AI-based systems, traditional tools are currently insufficient, and therefore data-driven methods should be used. These include simulation testing, testing systems in near-real-world conditions (HiL and Silent), and road tests, which can be conducted both in a controlled environment and in normal road traffic. This approach enables the assessment of vehicle performance across a wide range of situations, thereby increasing the system’s reliability and safety [289,290].

In summary, current research suggests that the future of autonomous freight transport will depend largely on the level of data collection and AI, rather than on hardware or road infrastructure. Therefore, intelligent sensor systems, data analysis, and AI-based operational models are becoming increasingly important. This approach makes road transport more flexible and better adapted to changing conditions. Moreover, solutions based on advanced machine learning and predictive algorithms improve system safety and efficiency and reduce the negative impact on the environment [291,292].

2.6. Advanced Intelligent Control and Coordination Systems in Autonomous Transport

The latest research directions in autonomous vehicles are increasingly less limited to sensor-based perception. There is growing emphasis on real-time decision-making, the development of intelligent control systems, and cooperative driving strategies.

One of the most important research directions is active safety mechanisms that enable autonomous vehicles to adapt their behaviour to sudden, unforeseen threats continuously. Such solutions combine real-time situation monitoring with corrective actions, allowing the vehicle to react even when higher-level planning fails. In practice, this translates into greater resilience of the entire system. Therefore, such approaches are now considered a key element in ensuring safety in complex road traffic environments.

Another rapidly developing area of research is cooperative platooning and the execution of coordinated maneuvers, which go beyond simply increasing fuel efficiency. Recent results suggest that such systems can also effectively handle more complex driving situations, such as synchronized intersection navigation and unprotected left turns. Communication, both vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I), plays a key role here. This enables better prediction of other road users’ behaviour and more coordinated real-time decision-making.

Furthermore, the development of machine learning and artificial intelligence has significantly enhanced the perception and decision-making capabilities of autonomous systems. AI-based models are increasingly being used to detect and classify objects, predict the movement trajectories of other road users, and develop adaptive driving strategies under uncertainty.

Thanks to this approach, autonomous vehicles can operate more effectively in mixed traffic environments, where they coexist with human-driven vehicles.

Despite these advances, the implementation of such intelligent control systems in real-world freight transport remains limited. Most research to date focuses on technical tests and simulation environments that allow for the verification of algorithms’ performance under controlled conditions. However, their actual impact on large-scale logistics, operating costs, and constraints imposed by existing infrastructure is significantly less frequently analyzed. In this context, this study complements the existing literature. It is important to emphasize that it combines a technological perspective with analyses of economic viability and infrastructure readiness. This work comprehensively presents the implementation of autonomous transport systems while reflecting the real-world operating conditions of this technology.

3. Empirical Research and Analytical Framework

3.1. Empirical Study on the Implementation of Autonomous Vehicles

Unlike studies that focus on a single area, this study employs a comprehensive research methodology that combines qualitative and quantitative analysis. The scope of the research includes a structured literature review, surveys of transport companies and road users, a strategic SWOT/TOWS analysis, and the development of a proprietary cost minimization model that accounts for infrastructure readiness. This approach provides a comprehensive assessment of the implementation possibilities and real-world impact of automation in freight transport.

To obtain reliable, accurate information on the impact of AVs on FT and to gather public opinion on their safety, surveys were conducted among TEes and among randomly selected individuals of various ages. The surveys of public opinion and TEes were supplemented by SWOT and TOWS analyses, from which conclusions were drawn, as well as a brief interview with the TE Wak Trans.

The surveys conducted among TEes aimed to assess knowledge, awareness, benefits, and risks of introducing AVs to the FT market. The collected responses also form the basis for calculating the benefits associated with reducing the number of drivers employed, reducing the vehicle fleet, and increasing mileage. Furthermore, companies were also asked about their awareness of the risks associated with AVs. In addition to the surveys aimed at TEes, people who use public roads daily also commented. The aim was to assess their sense of safety and awareness of AVs. The interview with Wak Trans provided a thorough insight into the car purchasing process and answered questions about the most important factors to consider when purchasing a car. The company also provided its opinion on the introduction of AVs and on the transport of goods using them.

For rigorous analysis of the empirical data, the study employed a set of statistical methods to verify relationships, differences, and the significance of observed patterns.

Pearson’s correlation analysis was used to identify and quantify linear relationships among key quantitative variables, such as the numbers of employees, drivers, and vehicles.

Spearman’s rank correlation was used as a nonparametric alternative to identify monotonic relationships when the assumption of normality was not fully met.

One-way analysis of variance (ANOVA) was also conducted to determine whether there were statistically significant differences between groups of companies classified by size or operational characteristics.

In the survey of 550 road users, additional statistical procedures were used to assess differences in perceptions and attitudes toward autonomous technologies. Student’s t-test was used to compare mean responses between two groups, for example, based on gender or driver’s licence status.

The Kruskal–Wallis test, a nonparametric equivalent of one-way analysis of variance (ANOVA), was used to determine differences among multiple groups, specifically age categories and levels of familiarity with autonomous vehicles.

The use of parametric and nonparametric methods increased the robustness of the analysis and allowed for a more comprehensive interpretation of the data.

Statistical procedures mitigate the weaknesses of descriptive analysis, providing a quantitative basis for assessing relationships and differences between groups, which strengthens the empirical nature of the study. However, due to the exploratory nature of the analysis and the sample size, the results should be interpreted with caution and treated primarily as support for the modelling and conceptual aspects of the study, rather than as fully generalizable conclusions.

The final stage of the research is to conduct a SWOT and TOWS analysis. A properly conducted analysis of all strengths, weaknesses, opportunities, and threats will provide an answer to the question of what impact AVs might have on TEes that rely on the transportation of all types of goods and merchandise.

All research and analyses converge on one goal: how AVs might impact FT. The results of the research are presented and analyzed below, and appropriate conclusions are drawn.

3.2. Comparative Analysis of Autonomous, Semi-Autonomous, and Conventional Vehicles

Regardless of whether the car is fully human-controlled, semi-autonomous, or autonomous, the type of power used to propel these vehicles remains the same. The same applies to the car’s power output. This has no impact on the degree of autonomy. However, the gearbox and automatic braking system do. Automatic gearbox and braking are mandatory components in AVs and SAVs. Comparison of technical data for conventional, semi-automatic and autonomous vehicles presented in

Table 1. Comparison of technical data.

Technical Data	Trucks
	Conventional	Semi-Automobile	Autonomous
Power Type	Petrol, Diesel, Gas, Hybrid, Hydrogen, Electric	Petrol, Diesel, Gas, Hybrid, Hydrogen, Electric	Petrol, Diesel, Gas, Hybrid, Hydrogen, Electric
Gearbox	Manual, Automatic	Automatic	Automatic
Handbrake	Manual, Electric (automatic)	Electric (automatic)	Electric (automatic)
Engine Power	Regardless of vehicle type	Regardless of vehicle type	Regardless of vehicle type

.

Table 2. Equipment comparison.

Equipment	Trucks
	Conventional	Semi-Automobile	Autonomous
Rearview camera	0	1	1
360-degree camera	0	1	1
Cruise control	0	0	0
Adaptive Cruise Control (ACC)	0	1	1
Lane keeping assist	0	1	1
GSM network access	0	1	0
5G network access	0	0	1
On-board computer	1	1	1
Head-up display	0	0	0

Legend: 0—Equipment not required for car operation, 1—Equipment required for car operation.

presents a comparison of the hardware and

Table 3. Comparison of security systems.

Security Systems	Trucks
	Conventional	Semi-Automobile	Autonomous
ABS	0 *	1	1
ESP	0 *	1	1
AEB	0	1	1
GPS	0	1	1
Radar	0	1	1
Lidar	0	0	1
Airbags	0 *	1	1
Reverse Object Detection	0	1	1
Driver Attention Monitoring	0	1	1
Automatic Speed Limiter	0	1	1

Legend: 0—System not required for vehicle operation, 1—System required for vehicle operation, *—System not required, but mandatory.

presents a comparison of the security systems.

Analyzing Table 2, which presents the equipment required for AVs and conventional vehicles, it can be concluded that traditional vehicles require only a basic computer to control the engine and electronics. In contrast, SAVs require a range of sensors and cameras, as well as access to a GSM network. Fully autonomous cars have the same equipment as SAVs, but additionally require access to a faster GSM network, which is why they use 5G. Of course, the control computer is much more advanced and processes much more data.

After analyzing the following table (Table 3), this time with safety systems, the conclusion is that no safety system is needed to drive a conventional car. However, airbags, ABS, and ESP are mandatory. The mandatory ABS was introduced in Poland on 1 July 2006. Airbags, however, have been mandatory since 2009. This is a requirement imposed by the European Union. The mandatory ESP system was introduced on 1 November 2014, for cars with a gross vehicle weight under 3.5 tons. In the case of autonomous cars, the list of safety systems is long, as they are the foundation for the vehicle’s proper functioning. Summarizing Table 1, Table 2 and Table 3, it is clear that no expensive and complex systems are needed to drive a conventional car, as the HD performs all the work and bears full responsibility for their actions. The only equipment required is a computer controlling the electronics and the engine, although this is not essential. Moving on to a semi-autonomous car (SAC), the chasm becomes clear. The difference in safety systems and equipment is significant. The on-board computer absorbs and processes large amounts of data collected from the environment to make appropriate decisions.

A fully autonomous car does not differ much from an SAC in terms of equipment, systems, and sensors. The difference occurs primarily in access to the GSM network, and not all SACs are equipped with lidar sensors. The most important thing here is a computer and appropriate software that can quickly process this vast amount of information and take appropriate action. It is also worth mentioning that external sensors in the infrastructure do not play a significant role in the proper operation of an SAC, as humans can still make decisions at any time. However, fully autonomous cars will not survive without them. A simplified structure of the neural network of the autonomous drive is shown in Figure 1.

3.3. Analysis of Survey Results from Transport Enterprises

The survey of TEes aimed to gather information on public awareness and safety regarding the use of AVs. It also examined the benefits and risks AVs could bring to companies, and how this would impact FT. The results of the study also helped obtain the data necessary for the calculations. The authors’ research results are preliminary. Statistically, they are not representative of the entire freight transport market and should not be used to draw general conclusions. The economic and operational implications presented in this work are based on the authors’ cost minimization model.

The empirical research consisted of two main stages, each aimed at gathering information regarding the feasibility of implementing autonomous and semi-autonomous technologies in transport companies.

The first stage of the pilot and exploratory research was conducted in January 2026. It covered a sample of 25 transport companies operating in Poland. Its primary goal was to identify key variables relevant to the implementation of automation. Specifically, attention was paid to parameters such as fleet size, employment structure, and operational patterns. This stage also served to determine the clarity and validity of the research tool used. As a result, this stage of the research aimed to provide preliminary information on the perceived benefits and risks of implementing autonomous vehicles. The results from this stage were used to refine the research tool, primarily by formulating questions, constructing response scales, and identifying key areas of analysis.

The second stage of the research was conducted in April 2026. Similar to the first stage, it again included 25 transport companies. The goal of this stage was to validate and deepen the findings obtained during the pilot phase. This research phase was intended to ensure the consistency and reliability of respondent responses and reduce the risks associated with random or one-off errors.

The two-stage research design enabled verification of respondents’ responses, thereby increasing the validity of the empirical results.

The empirical research conducted as part of this work was not intended to provide statistically representative conclusions for the entire national or European freight transport industry. The study aimed to support theoretical considerations and develop a cost minimization model. Furthermore, the research aimed to provide industry-specific, practical information necessary to identify the most important barriers and factors influencing the implementation of autonomous technologies in freight transport.

Therefore, the empirical research results should be considered preliminary and indicative. They supplement the analytical and modelling sections but do not allow generalization of the conclusions.

Analysis of the results of a study on the impact of AVs and SAVs on FT: The study involved 50 randomly selected, unaffiliated TEes operating in Poland. Questions 1, 3, and 6 were formulated to determine company size, the number of vehicles owned, and the number of employees and drivers employed. The size and level of employment in transport companies are shown in Figure 2.

The percentage distribution of company sizes in the Polish market is relatively even, with only large TEes with more than 26 employees having a larger distribution. This may be due to the companies’ location or the small size of the research group.

As with the number of employees, the number of vehicles owned is equal. Anomalies occur in the 6–15 vehicle range, perhaps due to short distances travelled or companies engaged in domestic transport. It may also be due to the rapid growth of TEes.

In Figure 2, Figure 3 and Figure 4 we can conclude that the relationship between employees, drivers, and vehicle ownership is proportional. This means that the larger the TE, the more vehicles, employees, and drivers it has. Hiring new employees and drivers goes hand in hand with company growth. This relates to the need for more drivers capable of transporting larger volumes of cargo. It can also be observed that smaller companies experience a faster increase in vehicle ownership than in employee ownership, indicating that the company is growing and there is a shortage of drivers in the market. These figures may change if autonomous or SAVs are introduced. Transportation costs may decrease, while company profits may increase.

The next question in the survey was intended to determine what types of goods are most frequently transported by Tees (Figure 5).

As can be seen, the most prominent companies are those involved in full-truckload transport of goods and food, followed by road transport, bulk goods, hazardous goods, and finally oversized transport. The responses obtained indicate that most TEes could implement AV transport. In food transport, this will be crucial, as it can significantly reduce shelf prices and shorten delivery times for perishables. An AV transporting other goods could also significantly reduce the final price. However, transporting dangerous and oversized goods using autonomous or SAVs is too risky and currently unfeasible (Figure 6).

Question 4 measured the average daily distance travelled by a tractor-trailer, and question 7 measured the average gross earnings of a driver in a TE. The answers are presented in

Table 4. Research results of 50 transport enterprises.

No. TEes	Average Gross Driver Salary [EUR] [E]	Average Daily Distance Travelled by One Tractor-Trailer [km] [F]	Benefits of Reducing the Vehicle Fleet [G]	Benefits of Increased Road Safety [HI]	Benefits of Reducing Transportation Costs [I]	Benefits of Reducing the Number of Drivers [J]	Threat of Loss of Communication with the Vehicle [K]	Threat of System Failure [L]	Threat of Sensor Damage [M]	Threat of Competition [NI]
1.	1190	800	4	2	3	3	4	5	4	3
2.	1666	600	4	3	4	2	5	5	4	3
3.	1547	500	3	1	3	2	1	3	3	4
4.	1696	570	1	1	1	1	1	1	1	1
5.	1428	300	5	5	1	1	5	5	5	5
6.	1904	800	5	3	2	1	2	1	5	5
7.	1280	550	5	5	5	5	5	1	5	3
8.	1904	400	1	1	1	1	4	4	4	3
9.	3095	700	1	4	1	5	5	5	5	5
10.	1309	650	3	2	2	1	5	5	4	5
11.	1904	600	3	3	5	3	5	5	4	3
12.	1428	400	2	3	5	3	4	5	4	4
13.	1904	890	1	1	1	1	5	5	5	5
14.	2142	550	3	5	5	4	5	5	4	2
15.	1785	500	2	2	5	5	5	4	3	1
16.	2023	650	5	3	5	5	5	5	5	4
17.	1309	450	5	4	5	4	5	5	5	5
18.	1500	600	4	5	5	5	5	5	4	5
19.	761	500	5	3	3	5	5	5	5	5
20.	1322	555	2	1	5	5	5	5	3	5
21.	1904	600	5	3	5	4	5	5	4	4
22.	2023	550	5	5	5	5	4	5	5	4
23.	2023	600	4	4	3	5	3	4	3	5
24.	1428	570	2	3	3	2	2	1	3	2
25.	1904	600	4	4	5	4	3	2	2	2
26.	1320	430	5	4	3	3	3	3	3	2
27.	1360	480	4	4	4	3	3	3	3	2
28.	1385	450	4	3	4	3	3	4	3	3
29.	1260	390	5	4	3	2	2	3	3	2
30.	1420	520	4	4	4	3	3	3	3	2
31.	1520	560	4	4	4	3	4	4	4	3
32.	1490	530	4	4	4	3	4	4	3	3
33.	1650	610	3	3	4	4	4	4	4	3
34.	1580	590	4	4	4	3	4	4	3	3
35.	1610	570	3	3	4	4	4	4	4	3
36.	1680	600	3	4	4	4	4	4	4	4
37.	1820	620	4	4	4	4	4	4	4	4
38.	1890	650	3	4	5	4	4	4	4	4
39.	1950	680	3	4	5	5	4	4	4	4
40.	2080	700	3	4	5	5	5	5	4	4
41.	1920	640	4	4	5	4	4	4	4	4
42.	2140	720	3	5	5	5	5	5	5	4
43.	2210	760	2	5	5	5	5	5	5	4
44.	1980	690	3	4	4	4	4	4	4	4
45.	2450	820	2	5	5	5	5	5	5	5
46.	2580	850	2	5	5	5	5	5	5	5
47.	2360	780	3	4	5	5	5	5	4	4
48.	2490	810	2	5	5	5	5	5	5	5
49.	2710	890	2	5	5	5	5	5	5	5
50.	2330	770	3	4	4	5	5	5	4	5
Average	1791.28	611.9	3.32	3.58	3.94	3.66	4.12	4.12	3.96	3.68
			Average 4 benefits	3.625			Average of 4 threats	3.97

.

Based on the average kilometres travelled by a single truck and the average earnings of a single driver, it can be assumed that implementing a single AV would save approximately EUR 1791.28 gross per month and cover twice as much distance per day, i.e., approximately 1200 km. As a result, workplaces will be re-educated, but the price of transported goods can be reduced by up to 10 percent.

Questions 9 and 10 constituted the most important part of the survey. They assessed the benefits and risks associated with introducing AVs into the company. Reviewing the results, it can be seen that the most significant benefits for companies are:

• Reducing transportation costs—rating 3.94;
• Reducing the vehicle fleet—rating 3.32;
• Reducing the number of drivers—rating 3.66, due to the fleet reduction;
• Reducing road safety—rating 3.58.

The average for all four benefit assessment graphs is:

$$\overline{x}={\displaystyle \frac{x_1+x_2+\cdots +x_N}{N}}={\displaystyle \frac{3.94+3.66+3.52+3.58}{4}}=3.625$$

The risk assessment was conducted in the same manner as the benefit assessment, and average scores were calculated for clarity and ease of comparison. The results are as follows:

• Loss of communication with the vehicle—score 4.12;
• Risk of system failure—score 4.12;
• Risk of sensor damage—score 3.96;
• Potential for competition to emerge—score 3.68.

The average for all four threat assessment graphs is:

$$\overline{x}={\displaystyle \frac{x_1+x_2+\cdots +x_N}{N}}={\displaystyle \frac{4.12+4.12+3.96+3.68}{4}}=3.97$$

Taking all the assessments and the calculated averages into account, it is clear that the threats perceived by TEes outweigh the benefits. In addition to closed-ended questions, the survey included open-ended questions that asked companies to discuss other benefits and threats not listed above. The most common responses regarding benefits concerned cost minimization, while the most common risks included loss of contact with the vehicle and the lack of appropriately adapted infrastructure.

To determine the strength and direction of the relationship between the studied variables, a correlation analysis was conducted using two coefficients: Spearman’s rho and Pearson’s r. These variables concern the functioning of the TEs and the perception of benefits and risks associated with the adopted solutions. Each analysis was based on a sample of 50 TEs observations. As a result, 91 pairs of variables were assessed (

Table 5. rho Spearman and Pearson R correlation results.

		rho Spearman Correlation Results				R-Pearson Correlation Results				rho Spearman Correlation Results				R-Pearson Correlation Results
No.	Pair of Variables	N	R	t(N-2)	p-Value	R-Pearson	p-Value	No.	Pair of Variables	N	R	t(N-2)	p-Value	R-Pearson	p-Value
1.	[A] & [B]	50	0.865210	11.95501	0.000000	0.8728	0.000	47.	[E] & [F]	50	0.676154	6.35829	0.000000	0.6568	0.000
2.	[A] & [C]	50	0.859472	11.64868	0.000000	0.8582	0.000	48.	[E] & [G]	50	−0.447874	−3.47050	0.001108	−0.4903	0.000
3.	[A] & [D]	50	0.449187	3.48324	0.001067	0.4477	0.001	49.	[E] & [HI]	50	0.389053	2.92596	0.005232	0.3528	0.012
4.	[A] & [E]	50	0.772752	8.43503	0.000000	0.7318	0.000	50.	[E] & [I]	50	0.358828	2.66341	0.010498	0.1635	0.257
5.	[A] & [F]	50	0.707486	6.93565	0.000000	0.6856	0.000	51.	[E] & [J]	50	0.507987	4.08589	0.000166	0.4226	0.002
6.	[A] & [G]	50	−0.206044	−1.45882	0.151128	−0.2050	0.153	52.	[E] & [K]	50	0.341965	2.52120	0.015071	0.3168	0.025
7.	[A] & [HI]	50	0.250695	1.79416	0.079087	0.2601	0.068	53.	[E] & [L]	50	0.357589	2.65286	0.010788	0.3031	0.032
8.	[A] & [I]	50	0.396467	2.99200	0.004367	0.2916	0.040	54.	[E] & [M]	50	0.398672	3.01177	0.004135	0.3589	0.010
9.	[A] & [J]	50	0.494999	3.94692	0.000258	0.4383	0.001	55.	[E] & [NI]	50	0.332950	2.44632	0.018147	0.3326	0.018
10.	[A] & [K]	50	0.455882	3.54865	0.000878	0.4245	0.002	56.	[F] & [G]	50	−0.368049	−2.74242	0.008546	−0.3750	0.007
11.	[A] & [L]	50	0.471831	3.70759	0.000542	0.3624	0.010	57.	[F] & [HI]	50	0.197499	1.39581	0.169196	0.1580	0.273
12.	[A] & [M]	50	0.410947	3.12301	0.003032	0.3665	0.009	58.	[F] & [I]	50	0.212553	1.50704	0.138353	0.1704	0.237
13.	[A] & [NI]	50	0.507982	4.08583	0.000166	0.4762	0.000	59.	[F] & [J]	50	0.386690	2.90506	0.005538	0.3361	0.017
14.	[B] & [C]	50	0.930665	17.62321	0.000000	0.9215	00.00	60.	[F] & [K]	50	0.366971	2.73313	0.008757	0.3169	0.025
15.	[B] & [D]	50	0.496796	3.96594	0.000243	0.4980	0.000	61.	[F] & [L]	50	0.369845	2.75792	0.008205	0.2323	0.104
16.	[B] & [E]	50	0.803425	9.34873	0.000000	0.7418	0.000	62.	[F] & [M]	50	0.462564	3.61469	0.000719	0.4071	0.003
17.	[B] & [F]	50	0.765692	8.24752	0.000000	0.7375	0.000	63.	[F] & [NI]	50	0.480198	3.79281	0.000417	0.4417	0.001
18.	[B] & [G]	50	−0.313514	−2.28741	0.026621	−0.2937	0.038	64.	[G] & [HI]	50	0.161013	1.13028	0.263976	0.3037	0.032
19.	[B] & [HI]	50	0.293986	2.13097	0.038241	0.3053	0.031	65.	[G] & [I]	50	−0.023497	−0.16284	0.871329	0.1668	0.247
20.	[B] & [I]	50	0.329718	2.41966	0.019372	0.2638	0.064	66.	[G] & [J]	50	−0.098965	−0.68903	0.494121	0.0053	0.971
21.	[B] & [J]	50	0.513001	4.14052	0.000139	0.4645	0.001	67.	[G] & [K]	50	−0.178647	−1.25794	0.214498	−0.0901	0.534
22.	[B] & [K]	50	0.351977	2.60529	0.012187	0.3427	0.015	68.	[G] & [L]	50	−0.177884	−1.25239	0.216494	−0.1255	0.385
23.	[B] & [L]	50	0.393562	2.96604	0.004690	0.3018	0.033	69.	[G] & [M]	50	−0.013507	−0.09359	0.925827	0.0489	0.736
24.	[B] & [M]	50	0.477341	3.76357	0.000456	0.4263	0.002	70.	[G] & [NI]	50	−0.099418	−0.69221	0.492138	−0.0413	0.776
25.	[B] & [NI]	50	0.500155	4.00165	0.000217	0.4916	0.000	71.	[HI] & [I]	50	0.448843	3.479903	0.001078	0.4819	0.000
26.	[C] & [D]	50	0.450468	3.49570	0.001028	0.4448	0.001	72.	[HI] & [J]	50	0.503404	4.036435	0.000194	0.5299	0.000
27.	[C] & [E]	50	0.811300	9.61439	0.000000	0.7784	0.000	73.	[HI] & [K]	50	0.265895	1.910963	0.061991	0.3181	0.024
28.	[C] & [F]	50	0.721785	7.22517	0.000000	0.6972	0.000	74.	[HI] & [L]	50	0.185156	1.305368	0.197990	0.1652	0.252
29.	[C] & [G]	50	−0.318305	−2.32628	0.024273	−0.2975	0.036	75.	[HI] & [M]	50	0.393991	2.969870	0.004641	0.3998	0.004
30.	[C] & [HI]	50	0.286692	2.07329	0.043535	0.2912	0.040	76.	[HI] & [NI]	50	0.212081	1.503545	0.139250	0.1938	0.177
31.	[C] & [I]	50	0.383115	2.87354	0.006030	0.2884	0.042	77.	[I] & [J]	50	0.661838	6.116682	0.000000	0.7168	0.000
32.	[C] & [J]	50	0.523374	4.25540	0.000096	0.4663	0.001	78.	[I] & [K]	50	0.441233	3.406485	0.001339	0.3762	0.007
33.	[C] & [K]	50	0.341677	2.51880	0.015162	0.3122	0.027	79.	[I] & [L]	50	0.345479	2.550599	0.013998	0.2688	0.059
34.	[C] & [L]	50	0.379112	2.83845	0.006626	0.3039	0.032	80.	[I] & [M]	50	0.213430	1.513563	0.136694	0.1370	0.343
35.	[C] & [M]	50	0.424211	3.24552	0.002139	0.3829	0.006	81.	[I] & [NI]	50	0.071109	0.493909	0.623623	0.0140	0.923
36.	[C] & [NI]	50	0.457874	3.56826	0.000827	0.4333	0.002	82.	[J] & [K]	50	0.519563	4.212904	0.000110	0.5117	0.000
37.	[D] & [E]	50	0.308617	2.247889	0.029214	0.2848	0.045	83.	[J] & [L]	50	0.407957	3.095731	0.003273	0.3862	0.006
38.	[D] & [F]	50	0.470576	3.694916	0.000563	0.4497	0.001	84.	[J] & [M]	50	0.368394	2.745395	0.008480	0.3167	0.025
39.	[D] & [G]	50	−0.019975	−0.138420	0.890487	−0.0106	0.942	85.	[J] & [NI]	50	0.376733	2.817681	0.007004	0.2857	0.044
40.	[D] & [HI]	50	0.174234	1.225877	0.226225	0.1919	0.182	86.	[K] & [L]	50	0.827137	10.19673	0.000000	0.7690	0.000
41.	[D] & [I]	50	0.196330	1.387215	0.171784	0.2404	0.093	87.	[K] & [M]	50	0.637892	5.73861	0.000001	0.6627	0.000
42.	[D] & [J]	50	0.279205	2.014504	0.049578	0.2916	0.040	88.	[K] & [NI]	50	0.504840	4.05187	0.000185	0.4752	0.000
43.	[D] & [K]	50	0.087287	0.607061	0.546671	0.1576	0.274	89.	[L] & [M]	50	0.594066	5.116512	0.000005	0.5180	0.000
44.	[D] & [L]	50	0.061346	0.425821	0.672141	0.0978	0.499	90.	[L] & [NI]	50	0.594739	5.125482	0.000005	0.5424	0.000
45.	[D] & [M]	50	0.138517	0.969015	0.337395	0.1466	0.310	91.	[M] & [NI]	50	0.641542	5.794301	0.000001	0.6843	0.000
46.	[D] & [NI]	50	0.249512	1.785127	0.080560	0.2714	0.057

).

The use of these two coefficients has significant methodological justification. The Spearman coefficient is used to assess the strength of the monotonic relationship between variables. It should be emphasized that it is also less sensitive to deviations from normality and outliers. The Pearson coefficient, on the other hand, allows for the assessment of the strength of the linear relationship between metric variables. Convergence between the results obtained with these two methods increases the reliability of interpretation. Discrepancies, on the other hand, may indicate a nonlinear nature of the relationship or a greater importance of rank order than strict linearity.

For reliable interpretation of the results, a statistical significance level of p < 0.05 was adopted. The relationships between variables describing the scale of TEes’ activities, operational parameters and perceived benefits and threats were analyzed.

The analysis revealed numerous positive, statistically significant correlations among variables describing the size and operation of TEes. Particularly strong correlations were observed between [A] and [B] (Spearman’s rho = 0.865210; Pearson’s r = 0.8728), as well as between [A] and [C] (rho = 0.859472; r = 0.8582). However, the strongest correlation across the entire dataset was found between [B] and [C] (rho = 0.930665; r = 0.9215). These results indicate very high structural coherence of the basic parameters of TEes’ operations: a larger employment scale is associated with a larger fleet and greater demand for drivers.

Strong positive correlations were also observed between economic and operational variables. [E] correlates positively with both [A] (rho = 0.772752; r = 0.7318), [B] (rho = 0.803425; r = 0.7418), and [C] (rho = 0.811300; r = 0.7784). Additionally, [F] remains strongly positively correlated with the same variables: with [A] (rho = 0.707486; r = 0.6856), [B] (rho = 0.765692; r = 0.7375), [C] (rho = 0.721785; r = 0.6972), as well as with [E] (rho = 0.676154; r = 0.6568). This means that larger TEs not only have larger fleets and employ more drivers, but also achieve higher operational and salary indicators.

[D] shows positive correlations, but significantly weaker ones than the basic structural variables. It is significantly related to [A] (rho = 0.449187; r = 0.4477), [B] (rho = 0.496796; r = 0.4980), [C] (rho = 0.450468; r = 0.4448), [E] (rho = 0.308617; r = 0.2848), and [F] (rho = 0.470576; r = 0.4497). This suggests that larger and more organizationally developed TEs demonstrate greater diversification of the cargo they transport, but this correlation is moderate rather than dominant.

Interesting results were obtained for variable [G]. In relation to the main parameters of TEes, this variable exhibits negative correlations, of which statistically significant are the relationships with [B] (rho = −0.313514; r = −0.2937), [C] (rho = −0.318305; r = −0.2975), [E] (rho = −0.447874; r = −0.4903), and [F] (rho = −0.368049; r = −0.3750). This means that the larger the scale of TEes’ operations and the higher the remuneration and intensity of fleet operation, the fewer perceived benefits of reducing the vehicle fleet. This can be interpreted as a reflection of the greater reliance of large TEes on maintaining transport capacity and greater caution regarding the reduction of transport resources.

In the area of perceived additional benefits, moderate positive correlations are visible. [HI] is positively correlated with, among others, [E] (rho = 0.389053; r = 0.3528), as well as [I] (rho = 0.448843; r = 0.4819) and [J] (rho = 0.503404; r = 0.5299). This indicates that TEs who perceive improved safety as an important benefit are more likely to perceive cost and organizational benefits simultaneously.

The relationships between variables related to economic and personnel benefits are particularly important. The strongest relationship in this part of the model was found between [I] and [J] (rho = 0.661838; r = 0.7168). This result suggests that both benefit dimensions are perceived by TEs as strongly co-occurring. [A] (rho = 0.494999; r = 0.4383), [B] (rho = 0.513001; r = 0.4645), [C] (rho = 0.523374; r = 0.4663), [E] (rho = 0.507987; r = 0.4226), and [F] (rho = 0.386690; r = 0.3361) are also positively associated with [J]. This indicates that larger entities are more likely to perceive the potential benefits of reducing the demand for driver labour, which may be due to larger staffing and cost issues.

A very clear and consistent pattern of relationships also appears among the variables describing threats. [K] is very strongly related to [L] (rho = 0.827137; r = 0.7690), as well as to [M] (rho = 0.637892; r = 0.6627) and [NI] (rho = 0.504840; r = 0.4752). In turn, [L] positively correlates with [M] (rho = 0.594066; r = 0.5180) and with [NI] (rho = 0.594739; r = 0.5424), and [M] remains strongly related to [NI] (rho = 0.641542; r = 0.6843). These results suggest that TEes perceive technical and market risks in an interconnected manner, forming a relatively coherent dimension of risk perception.

It should also be noted that some benefit-related variables are positively correlated with perceived risks. For example, [J] correlates positively with [K] (rho = 0.519563; r = 0.5117), [L] (rho = 0.407957; r = 0.3862), [M] (rho = 0.368394; r = 0.3167), and [NI] (rho = 0.376733; r = 0.2857). This means that TEs who perceive the potential benefits of transport transformation more strongly are also more likely to identify the associated technological and market risks. According to the authors, this does not indicate a contradiction in attitudes but rather a more in-depth and realistic assessment of the analyzed phenomenon.

Comparing the Spearman and Pearson results, one can observe a high degree of consistency in direction and similar strength of correlation for most pairs of variables. This consistency allows us to conclude that the analyzed relationships are most often both monotonic and close to linear. This is particularly evident in the case of the strongest correlations, such as [B] and [C], [A] and [B], [A] and [C], and the relationship between [K] and [L].

It is also important to note that there are individual discrepancies between the results of the two coefficients. For example, the relationship between [E] and [I] is significant for Spearman’s rho (0.358828; p = 0.010498) but insignificant for Pearson’s r (0.1635; p = 0.257). Similarly, the relationship between [F] and [L] is significant for Spearman’s rho (0.369845; p = 0.008205) but not for Pearson’s rho (0.2323; p = 0.104). The opposite situation is observed for the pair [G] and [HI], where significance is observed only for Pearson’s rho. Such differences suggest that some of the relationships may deviate from complete linearity. Furthermore, they may also result from the influence of the rank order or the specificity of the data distribution.

Based on the correlation analysis, it can be concluded that the variables studied constitute a coherent structure of relationships, consisting of several distinct areas of interrelationship. First, the basic characteristics describing the scale of TEes’ operations, such as [A], [B], and [C], are very strongly and positively correlated. Consequently, an increase in one variable leads to increases in the other variables. This confirms the structural interdependence of these variables in the TEes’ operational model. Second, the economic and operational variables [E] and [F] are strongly correlated with [A]. This suggests that larger entities exhibit more intensive resource use and higher labour costs. This may be due to the scale and complexity of their operations. Third, variables [G] are less considered as significant in larger and more intensively operating TEes. Negative correlations in this area indicate that fleet reduction is not perceived as beneficial in entities where transport capacity serves as a key strategic resource. Fourth, the perception of technological and market threats creates a clearly coherent pattern of relationships. The particularly strong correlations between [K], [L], [M], and [NI] indicate that TEs treat these risks as elements of a single, common area of uncertainty. Fifth, the results obtained using Spearman and Pearson coefficients are largely consistent. This confirms the stability of the observed relationships and increases the confidence in their interpretation. At the same time, individual differences between the two measures emerge. This suggests that some relationships may not be fully linear. This is worth considering in further analyses. In summary, the results indicate that in the studied TEs sample, there are clear correlations among the scale of operations, operational parameters, and perceptions of benefits and threats. This pattern of relationships suggests that larger TEs assess the effects of organizational and technological changes differently than smaller entities. Furthermore, they also demonstrate a greater perception of economic benefits and technical threats.

One-way analysis of variance was used to determine whether the mean values of the studied variables differed significantly across the categories identified as [A], [B], [C], [D], and [P]. ANOVA allows determining whether a main effect of a factor exists between the compared groups, while the post hoc Fisher’s LSD test indicates which specific pairs of groups differ statistically significantly. A significance criterion of p < 0.05 was adopted for interpretation.

Methodologically, this means that the overall mean differences between the groups were first assessed, and the source of these differences was only identified in a second step. It should be noted that Fisher’s LSD test is relatively liberal, so borderline results should be interpreted with caution, especially with a large number of comparisons. Nevertheless, the combination of ANOVA and post hoc tests provides a clear understanding of the structure of differences among the studied TEs (

Table 6. One-way ANOVA and Fisher’s LSD test results—Size and level of employment in a transport enterprises [A].

	One-Way ANOVA											Test Post Hoc—Fisher’s LSD Test
	Variable	Fisher-Snedecor Test	p-Value	[A1]		[A2]		[A3]		[A4]		Pair of Variables
												[A1] & [A2]	[A1] & [A3]	[A1] & [A4]	[A2] & [A3]	[A2] & [A4]
				Mean	Standard Deviation	Mean	Standard Deviation	Mean	Standard Deviation	Mean	Standard Deviation	p-Value	p-Value	p-Value	p-Value	p-Value
1.	[A]
2.	[B]	58.714	0.000	1.000	0.000	2.417	0.515	2.833	0.835	3.813	0.403	0.000	0.000	0.000	0.061	0.000
3.	[C]	44.161	0.000	1.300	0.483	1.917	0.669	2.917	0.793	3.813	0.403	0.020	0.000	0.000	0.000	0.000
4.	[D]	4.039	0.012	1.700	0.483	2.333	1.155	2.500	1.000	3.063	1.063	0.140	0.064	0.001	0.681	0.059
5.	[P1]	4.492	0.008	0.000	0.000	0.000	0.000	0.250	0.452	0.438	0.512	1.000	0.118	0.005	0.102	0.003
6.	[P2]	8.445	0.000	0.200	0.422	0.417	0.515	0.250	0.452	0.438	0.512	0.316	0.316	0.000	1.000	0.001
7.	[P3]	0.969	0.416	0.100	0.316	0.333	0.492	0.417	0.515	0.375	0.500
8.	[P4]	0.739	0.534	0.600	0.516	0.667	0.492	0.333	0.492	0.000	0.000
9.	[P5]	7.508	0.000	0.400	0.516	0.250	0.452	0.500	0.522	0.533	0.516	0.706	0.136	0.001	0.052	0.000
10.	[P6]	0.813	0.494	0.400	0.516	0.333	0.492	0.500	0.522	0.625	0.500
11.	[P7]	0.861	0.468	1.700	0.483	2.333	1.155	2.500	1.000	3.063	1.063
12.	[F]	13.839	0.000	482.000	133.150	542.083	63.010	646.667	72.530	719.375	118.742	0.173	0.000	0.000	0.015	0.000
13.	[E]	20.460	0.000	1409.400	200.624	1456.167	254.757	1907.667	248.087	2194.000	401.341	0.718	0.000	0.000	0.001	0.000
14.	[G]	1.340	0.273	4.000	1.247	3.167	1.267	3.083	0.669	3.188	1.424
15.	[HI]	1.581	0.207	3.300	1.494	3.083	1.084	3.833	1.030	3.938	1.124
16.	[I]	2.473	0.073	3.100	1.287	3.833	1.115	4.500	0.905	4.125	1.500
17.	[J]	3.900	0.015	2.600	1.174	3.417	1.165	4.000	1.279	4.250	1.390	0.140	0.013	0.002	0.267	0.093
18.	[K]	4.137	0.011	3.300	1.252	3.833	1.193	4.583	0.515	4.500	0.966	0.223	0.005	0.005	0.075	0.090
19.	[L]	2.651	0.060	3.400	1.174	3.833	1.403	4.500	0.522	4.500	1.211
20.	[M]	2.679	0.058	3.600	0.843	3.583	1.084	4.083	0.515	4.375	0.885
21.	[NI]	4.905	0.005	2.900	0.994	3.417	1.240	3.583	1.084	4.438	0.892	0.256	0.135	0.001	0.699	0.014

,

Table 7. One-way ANOVA and Fisher’s LSD test results—Number of trucks owned [B].

	One-Way ANOVA											Test Post Hoc—Fisher’s LSD Test
	Variable	Fisher-Snedecor Test	p-Value	[B1]		[B2]		[B3]		[B4]		Pair of Variables
												[B1] & [B2]	[B1] & [B3]	[B1] & [B4]	[B2] & [B3]	[B2] & [B4]	[B3] & [B4]
				Mean	Standard Deviation	Mean	Standard Deviation	Mean	Standard Deviation	Mean	Standard Deviation	p-Value	p-Value	p-Value	p-Value	p-Value	p-Value
1.	[A]	58.714	0.000	1.000	0.000	2.417	0.515	2.833	0.835	3.813	0.403	0.000	0.000	0.000	0.061	0.000	0.000
2.	[B]
3.	[C]	106.160	0.000	1.300	0.483	1.833	0.577	2.750	0.452	4.000	0.000	0.005	0.000	0.000	0.000	0.000	0.000
4.	[D]	6.946	0.001	1.700	0.483	1.917	0.900	3.083	1.084	2.938	0.998	0.585	0.001	0.002	0.003	0.006	0.680
5.	[P1]	4.492	0.008	0.000	0.000	0.000	0.000	0.250	0.452	0.438	0.512	1.000	0.118	0.005	0.102	0.003	0.187
6.	[P2]	11.042	0.000	0.000	0.000	0.000	0.000	0.333	0.492	0.688	0.479	1.000	0.038	0.000	0.030	0.000	0.014
7.	[P3]	1.518	0.222	0.100	0.316	0.250	0.452	0.500	0.522	0.375	0.500
8.	[P4]	1.561	0.212	0.200	0.422	0.333	0.492	0.583	0.515	0.250	0.447
9.	[P5]	15.476	0.000	0.600	0.516	0.833	0.389	0.167	0.389	0.000	0.000	0.130	0.006	0.000	0.000	0.000	0.223
10.	[P6]	0.813	0.494	0.400	0.516	0.250	0.452	0.500	0.522	0.533	0.516
11.	[P7]	1.576	0.208	0.400	0.516	0.250	0.452	0.583	0.515	0.625	0.500
12.	[F]	19.499	0.000	482.000	133.150	540.417	70.532	617.500	41.369	742.500	104.147	0.148	0.001	0.000	0.048	0.000	0.001
13.	[E]	22.942	0.000	1409.400	200.624	1520.667	345.502	1758.000	207.506	2257.875	341.562	0.377	0.008	0.000	0.052	0.000	0.000
14.	[G]	1.881	0.146	4.000	1.247	3.250	1.288	3.417	0.793	2.875	1.310
15.	[HI]	2.225	0.098	3.300	1.494	3.000	1.279	3.750	0.622	4.063	1.124
16.	[I]	2.365	0.083	3.100	1.287	3.917	1.311	4.500	0.674	4.063	1.482
17.	[J]	4.381	0.009	2.600	1.174	3.333	1.435	4.000	0.853	4.313	1.401	0.179	0.012	0.001	0.200	0.047	0.518
18.	[K]	3.330	0.028	3.300	1.252	4.333	1.155	4.000	0.853	4.563	0.892	0.023	0.119	0.004	0.432	0.563	0.159
19.	[L]	2.831	0.049	3.400	1.174	4.333	1.155	3.833	1.193	4.625	1.025	0.059	0.374	0.010	0.283	0.502	0.072
20.	[M]	5.428	0.003	3.600	0.843	3.667	1.073	3.667	0.651	4.625	0.619	0.847	0.847	0.003	1.000	0.003	0.003
21.	[NI]	5.674	0.002	2.900	0.994	3.417	1.505	3.500	0.905	4.500	0.632	0.247	0.180	0.000	0.844	0.008	0.014

,

Table 8. One-way ANOVA and Fisher’s LSD test results—Drivers employed in transport enterprises [C].

	One-Way ANOVA											Test Post Hoc—Fisher’s LSD Test
	Variable	Fisher-Snedecor Test	p-Value	[C1]		[C2]		[C3]		[C4]		Pair of Variables
												[C1] & [C2]	[C1] & [C3]	[C1] & [C4]	[C2] & [C3]	[C2] & [C4]	[C3] & [C4]
				Mean	Standard Deviation	Mean	Standard Deviation	Mean	Standard Deviation	Mean	Standard Deviation	p-Value	p-Value	p-Value	p-Value	p-Value	p-Value
1.	[A]	43.579	0.000	1.300	0.483	2.071	0.730	3.100	0.738	3.813	0.403	0.003	0.000	0.000	0.000	0.000	0.005
2.	[B]	91.796	0.000	1.300	0.483	2.000	0.679	2.900	0.316	4.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000
3.	[C]
4.	[D]	5.452	0.003	1.900	0.876	1.929	0.829	3.100	1.101	2.938	0.998	0.943	0.007	0.010	0.005	0.006	0.674
5.	[P1]	4.984	0.004	0.000	0.000	0.000	0.000	0.300	0.483	0.438	0.512	1.000	0.071	0.004	0.051	0.002	0.351
6.	[P2]	9.566	0.000	0.000	0.000	0.071	0.267	0.300	0.483	0.688	0.479	0.648	0.080	0.000	0.148	0.000	0.014
7.	[P3]	1.011	0.397	0.200	0.422	0.214	0.426	0.500	0.527	0.375	0.500
8.	[P4]	1.007	0.398	0.200	0.422	0.429	0.514	0.500	0.527	0.250	0.447
9.	[P5]	10.214	0.000	0.700	0.483	0.643	0.497	0.200	0.422	0.000	0.000	0.723	0.006	0.000	0.008	0.000	0.207
10.	[P6]	1.528	0.220	0.400	0.516	0.214	0.426	0.600	0.516	0.533	0.516
11.	[P7]	0.796	0.503	0.400	0.516	0.357	0.497	0.500	0.527	0.625	0.500
12.	[F]	16.402	0.000	505.000	129.979	537.500	83.315	614.000	57.194	742.500	104.147	0.423	0.016	0.000	0.064	0.000	0.002
13.	[E]	26.238	0.000	1380.000	292.448	1529.643	246.416	1822.300	176.630	2257.875	341.562	0.202	0.001	0.000	0.015	0.000	0.000
14.	[G]	1.857	0.150	4.000	1.247	3.286	1.204	3.400	0.843	2.875	1.310
15.	[HI]	1.788	0.163	3.300	1.494	3.143	1.167	3.700	0.823	4.063	1.124
16.	[I]	3.398	0.025	3.100	1.287	3.786	1.188	4.800	0.422	4.063	1.482	0.179	0.003	0.055	0.049	0.536	0.139
17.	[J]	5.791	0.002	2.900	1.370	2.929	1.207	4.400	0.516	4.313	1.401	0.955	0.008	0.006	0.005	0.003	0.859
18.	[K]	1.982	0.130	3.800	1.229	3.714	1.326	4.300	0.675	4.563	0.892
19.	[L]	1.673	0.186	3.700	1.252	3.857	1.406	4.100	0.876	4.625	1.025
20.	[M]	6.389	0.001	3.900	0.876	3.429	0.938	3.700	0.675	4.625	0.619	0.153	0.571	0.026	0.407	0.000	0.005
21.	[NI]	5.048	0.004	3.300	1.059	3.143	1.292	3.500	1.179	4.500	0.632	0.718	0.671	0.007	0.414	0.001	0.022

,

Table 9. One-way ANOVA and Fisher’s LSD test results—Number of types of goods transported [D].

	One-Way ANOVA											Test Post Hoc—Fisher’s LSD Test
	Variable	Fisher-Snedecor Test	p-Value	[D1]		[D2]		[D3]		[D4]		Pair of Variables
												[D1] & [D2]	[D1] & [D3]	[D1] & [D4]	[D2] & [D3]	[D2] & [D4]	[D3] & [D4]
				Mean	Standard Deviation	Mean	Standard Deviation	Mean	Standard Deviation	Mean	Standard Deviation	p-Value	p-Value	p-Value	p-Value	p-Value	p-Value
1.	[A]	6.524	0.001	2.300	1.160	2.056	1.056	3.400	0.699	3.333	0.888	0.530	0.016	0.018	0.001	0.001	0.874
2.	[B]	9.602	0.000	2.200	1.135	2.000	1.029	3.600	0.699	3.333	0.651	0.583	0.001	0.006	0.000	0.000	0.500
3.	[C]	8.042	0.000	2.200	1.135	2.000	0.970	3.600	0.699	3.167	0.937	0.597	0.002	0.022	0.000	0.002	0.293
4.	[D]
5.	[P1]	3.268	0.030	538.000	178.126	565.833	96.927	679.000	95.621	686.667	123.091	0.767	0.022	0.360	0.005	0.175	0.130
6.	[P2]	11.878	0.000	0.300	0.483	0.056	0.236	0.400	0.516	0.667	0.492	0.483	0.001	0.003	0.000	0.000	0.451
7.	[P3]	5.234	0.003	0.100	0.316	0.056	0.236	0.500	0.527	0.250	0.452	0.147	0.597	0.047	0.043	0.000	0.145
8.	[P4]	4.291	0.009	0.100	0.316	0.000	0.000	0.700	0.483	0.583	0.515	0.059	0.131	0.001	0.847	0.046	0.056
9.	[P5]	5.048	0.004	0.000	0.000	0.333	0.485	0.300	0.483	0.667	0.492	0.038	0.308	0.477	0.002	0.003	0.721
10.	[P6]	2.161	0.106	0.200	0.422	0.333	0.485	0.556	0.527	0.667	0.492
11.	[P7]	5.479	0.003	0.000	0.000	0.556	0.511	0.500	0.527	0.750	0.452	0.003	0.016	0.000	0.754	0.249	0.198
12.	[F]	4.526	0.007	1747.900	547.290	1576.111	437.184	2028.300	226.052	1952.667	381.547	0.568	0.014	0.007	0.024	0.011	0.885
13.	[E]	3.307	0.028	3.100	1.595	3.444	1.149	3.600	1.174	3.083	1.084	0.302	0.140	0.258	0.009	0.019	0.674
14.	[G]	0.481	0.697	3.500	1.509	3.167	1.150	4.100	0.876	3.833	1.115
15.	[HI]	1.595	0.203	3.400	1.838	3.833	1.150	4.300	1.059	4.250	1.138
16.	[I]	1.106	0.357	3.100	1.729	3.333	1.237	4.400	1.265	4.000	1.128
17.	[J]	2.248	0.095	4.000	1.414	3.944	1.211	4.200	1.033	4.417	0.669
18.	[K]	0.485	0.694	4.000	1.414	3.944	1.211	4.200	1.033	4.417	0.669
19.	[L]	0.553	0.648	4.200	1.317	3.833	1.295	4.300	1.252	4.333	0.888
20.	[M]	1.958	0.134	3.900	1.197	3.667	0.767	4.500	0.707	4.000	0.853
21.	[NI]	2.309	0.089	3.500	1.434	3.222	1.215	4.200	0.632	4.083	0.996

,

Table 10. One-way ANOVA and Fisher’s LSD test results—Dangerous Goods [P1] and Oversized Transport [P2].

	[P1]								[P2]
	One-Way ANOVA							Test Post Hoc—Fisher’s LSD Test	One-Way ANOVA						Test Post Hoc—Fisher’s LSD Test
	Variable	Fisher-Snedecor Test	p-Value	[No]		[Yes]		[No] & [Yes]	Fisher-Snedecor Test	p-Value	[No]		[Yes]		[No] & [Yes]
				Mean	Standard Deviation	Mean	Standard Deviation	p-Value			Mean	Standard Deviation	Mean	Standard Deviation	p-Value
1.	[A]	12.516	0.001	2.425	1.107	3.700	0.483	0.001	19.457	0.000	2.286	1.045	3.600	0.737	0.000
2.	[B]	12.516	0.001	2.425	1.107	3.700	0.483	0.001	29.183	0.000	2.229	1.031	3.733	0.458	0.000
3.	[C]	13.626	0.001	2.375	1.102	3.700	0.483	0.001	26.486	0.000	2.200	1.023	3.667	0.617	0.000
4.	[D]	3.055	0.087	2.350	1.075	3.000	0.943		18.323	0.000	2.114	0.963	3.333	0.816	0.000
5.	[P1]								2.400	0.128	0.143	0.355	0.333	0.488
6.	[P2]	2.400	0.128	0.250	0.439	0.500	0.527
7.	[P3]	2.826	0.099	0.375	0.490	0.100	0.316		4.726	0.035	0.229	0.426	0.533	0.516	0.035
8.	[P4]	0.086	0.771	0.350	0.483	0.300	0.483		1.517	0.224	0.286	0.458	0.467	0.516
9.	[P5]	7.855	0.007	0.450	0.504	0.000	0.000	0.007	15.247	0.000	0.514	0.507	0.000	0.000	0.000
10.	[P6]	0.317	0.576	0.400	0.496	0.500	0.527		0.034	0.855	0.429	0.502	0.400	0.507
11.	[P7]	0.702	0.406	0.450	0.504	0.600	0.516		1.217	0.275	0.429	0.502	0.600	0.507
12.	[E]	15.667	0.000	1681.850	393.079	2229.000	381.778	0.000	8.453	0.006	1679.514	444.436	2052.067	333.763	0.006
13.	[F]	9.106	0.004	585.125	128.090	719.000	113.475	0.004	13.071	0.001	571.286	123.404	706.667	116.169	0.001
14.	[G]	2.334	0.133	3.450	1.131	2.800	1.476		0.003	0.960	3.314	1.207	3.333	1.291
15.	[HI]	2.429	0.126	3.450	1.154	4.100	1.287		1.235	0.272	3.457	1.197	3.867	1.187
16.	[I]	0.495	0.485	3.875	1.202	4.200	1.687		0.854	0.360	3.829	1.317	4.200	1.265
17.	[J]	3.804	0.057	3.475	1.358	4.400	1.265		4.436	0.040	3.400	1.311	4.267	1.387	0.040
18.	[K]	3.665	0.062	3.975	1.143	4.700	0.675		2.180	0.146	3.971	1.150	4.467	0.915
19.	[L]	3.102	0.085	3.975	1.209	4.700	0.949		2.678	0.108	3.943	1.211	4.533	1.060
20.	[M]	4.823	0.033	3.825	0.844	4.500	0.972	0.033	7.671	0.008	3.743	0.919	4.467	0.640	0.008
21.	[NI]	2.555	0.117	3.550	1.197	4.200	0.919		7.605	0.008	3.400	1.218	4.333	0.724	0.008

,

Table 11. One-way ANOVA and Fisher’s LSD test results—Bulk Goods [P3] and Collective Freight Transport [P4].

	[P3]								[P4]
	One-Way ANOVA							Test Post Hoc—Fisher’s LSD Test	One-Way ANOVA						Test Post Hoc—Fisher’s LSD Test
	Variable	Fisher-Snedecor Test	p-Value	[No]		[Yes]		[No] & [Yes]	Fisher-Snedecor Test	p-Value	[No]		[Yes]		[No] & [Yes]
				Mean	Standard Deviation	Mean	Standard Deviation	p-Value			Mean	Standard Deviation	Mean	Standard Deviation	p-Value
1.	[A]	1.913	0.173	2.529	1.187	3.000	0.966		0.819	0.370	2.576	1.146	2.882	1.111
2.	[B]	2.780	0.102	2.500	1.187	3.063	0.929		0.141	0.708	2.636	1.220	2.765	0.970
3.	[C]	1.627	0.208	2.500	1.161	2.938	1.063		0.001	0.975	2.636	1.220	2.647	0.996
4.	[D]	7.902	0.007	2.206	0.914	3.063	1.181	0.007	10.927	0.002	2.152	1.004	3.118	0.928	0.002
5.	[P1]	2.826	0.099	0.265	0.448	0.063	0.250		0.086	0.771	0.212	0.415	0.176	0.393
6.	[P2]	4.726	0.035	0.206	0.410	0.500	0.516	0.035	1.517	0.224	0.242	0.435	0.412	0.507
7.	[P3]								0.976	0.328	0.273	0.452	0.412	0.507
8.	[P4]	0.976	0.328	0.294	0.462	0.438	0.512
9.	[P5]	10.592	0.002	0.500	0.508	0.063	0.250	0.002	0.005	0.942	0.364	0.489	0.353	0.493
10.	[P6]	1.096	0.300	0.471	0.507	0.313	0.479		0.007	0.934	0.424	0.502	0.412	0.507
11.	[P7]	1.981	0.166	0.412	0.500	0.625	0.500		1.654	0.205	0.545	0.506	0.353	0.493
12.	[E]	1.951	0.169	1731.471	468.196	1918.375	375.837		0.852	0.361	1833.091	509.322	1710.118	280.844
13.	[F]	1.362	0.249	596.618	137.965	644.375	128.113		0.177	0.676	617.727	149.515	600.588	106.211
14.	[G]	0.274	0.603	3.382	1.371	3.188	0.834		1.252	0.269	3.182	1.286	3.588	1.064
15.	[HI]	0.886	0.351	3.471	1.261	3.813	1.047		0.045	0.833	3.606	1.345	3.529	0.874
16.	[I]	1.347	0.251	3.794	1.431	4.250	0.931		0.054	0.818	3.909	1.422	4.000	1.061
17.	[J]	0.951	0.334	3.529	1.461	3.938	1.181		0.227	0.636	3.727	1.506	3.529	1.125
18.	[K]	0.276	0.602	4.176	1.086	4.000	1.155		0.078	0.781	4.152	1.176	4.059	0.966
19.	[L]	0.074	0.786	4.088	1.264	4.188	1.047		0.258	0.614	4.182	1.211	4.000	1.173
20.	[M]	0.045	0.832	3.941	1.013	4.000	0.632		2.087	0.155	4.091	0.947	3.706	0.772
21.	[NI]	2.605	0.113	3.500	1.237	4.063	0.929		0.423	0.519	3.758	1.173	3.529	1.179

,

Table 12. One-way ANOVA and Fisher’s LSD test results—Food [P5] and Car Transport [P6].

	[P5]								[P6]
	One-Way ANOVA							Test Post Hoc—Fisher’s LSD Test	One-Way ANOVA						Test Post Hoc—Fisher’s LSD Test
	Variable	Fisher-Snedecor Test	p-Value	[No]		[Yes]		[No] & [Yes]	Fisher-Snedecor Test	p-Value	[No]		[Yes]		[No] & [Yes]
				Mean	Standard Deviation	Mean	Standard Deviation	p-Value			Mean	Standard Deviation	Mean	Standard Deviation	p-Value
1.	[A]	18.661	0.000	3.125	1.070	1.889	0.758	0.000	0.883	0.352	2.552	1.121	2.857	1.153
2.	[B]	27.482	0.000	3.188	1.030	1.778	0.647	0.000	0.883	0.352	2.552	1.121	2.857	1.153
3.	[C]	28.550	0.000	3.156	1.019	1.722	0.669	0.000	1.326	0.255	2.483	1.122	2.857	1.153
4.	[D]	3.489	0.068	2.688	1.148	2.111	0.832		6.278	0.016	2.172	1.002	2.905	1.044	0.016
5.	[P1]	7.855	0.007	0.313	0.471	0.000	0.000	0.007	0.317	0.576	0.172	0.384	0.238	0.436
6.	[P2]	15.247	0.000	0.469	0.507	0.000	0.000	0.000	0.034	0.855	0.310	0.471	0.286	0.463
7.	[P3]	10.592	0.002	0.469	0.507	0.056	0.236	0.002	1.096	0.300	0.379	0.494	0.238	0.436
8.	[P4]	0.005	0.942	0.344	0.483	0.333	0.485		0.007	0.934	0.345	0.484	0.333	0.483
9.	[P5]								2.352	0.132	0.448	0.506	0.238	0.436
10.	[P6]	2.352	0.132	0.500	0.508	0.278	0.461
11.	[P7]	2.445	0.124	0.563	0.504	0.333	0.485		0.371	0.545	0.517	0.509	0.429	0.507
12.	[E]	28.670	0.000	1993.625	400.703	1431.556	256.199	0.000	1.713	0.197	1721.586	486.502	1887.524	372.167
13.	[F]	6.603	0.013	646.875	144.209	549.722	92.585	0.013	0.004	0.950	612.931	134.427	610.476	140.195
14.	[G]	4.223	0.045	3.063	1.190	3.778	1.166	0.045	0.028	0.868	3.345	1.203	3.286	1.271
15.	[HI]	1.824	0.183	3.750	1.191	3.278	1.179		1.981	0.166	3.379	1.237	3.857	1.108
16.	[I]	0.786	0.380	4.063	1.413	3.722	1.074		2.646	0.110	3.690	1.312	4.286	1.231
17.	[J]	2.213	0.143	3.875	1.408	3.278	1.274		1.143	0.290	3.483	1.430	3.905	1.300
18.	[K]	0.712	0.403	4.219	1.070	3.944	1.162		0.819	0.370	4.000	1.254	4.286	0.845
19.	[L]	2.396	0.128	4.313	1.120	3.778	1.263		0.132	0.718	4.172	1.167	4.048	1.244
20.	[M]	6.252	0.016	4.188	0.780	3.556	0.984	0.016	0.810	0.373	3.862	0.915	4.095	0.889
21.	[NI]	7.562	0.008	4.000	0.950	3.111	1.323	0.008	0.097	0.757	3.724	1.251	3.619	1.071

and

Table 13. One-way ANOVA and Fisher’s LSD test results—Full Truck Load Transport [P7].

	[P7]
	One-Way ANOVA							Test Post Hoc—Fisher’s LSD Test
	Variable	Fisher-Snedecor Test	p-Value	[No]		[Yes]		[No] & [Yes]
				Mean	Standard Deviation	Mean	Standard Deviation	p-Value
1.	[A]	2.058	0.158	2.462	1.104	2.917	1.139
2.	[B]	2.894	0.095	2.423	1.102	2.958	1.122
3.	[C]	2.007	0.163	2.423	1.102	2.875	1.154
4.	[P1]	0.702	0.406	0.154	0.368	0.250	0.442
5.	[P2]	1.217	0.275	0.231	0.430	0.375	0.495
6.	[P3]	1.981	0.166	0.231	0.430	0.417	0.504
7.	[P4]	1.654	0.205	0.423	0.504	0.250	0.442
8.	[P5]	2.445	0.124	0.462	0.508	0.250	0.442
9.	[P6]	0.535	0.468	0.480	0.510	0.375	0.495
10.	[P7]
11.	[D]	11.038	0.002	2.038	1.038	2.958	0.908	0.034
12.	[E]	2.034	0.160	1705.808	398.560	1883.875	483.005
13.	[F]	11.630	0.001	555.000	122.776	673.542	122.816	0.001
14.	[G]	1.184	0.282	3.500	1.334	3.125	1.076
15.	[HI]	0.203	0.654	3.654	1.129	3.500	1.285
16.	[I]	0.009	0.925	3.923	1.324	3.958	1.301
17.	[J]	0.416	0.522	3.538	1.363	3.792	1.414
18.	[K]	0.082	0.776	4.077	1.129	4.167	1.090
19.	[L]	0.547	0.463	4.000	1.296	4.250	1.073
20.	[M]	0.373	0.544	3.885	0.952	4.042	0.859
21.	[N]	4.763	0.034	3.346	1.231	4.042	0.999	0.544

).

Variable [A] (Table 5) significantly differentiated variables characterizing the scale of operation and basic operational parameters of the studied units. The strongest effects were observed for variables: [B] (F = 58.714; p < 0.001), [C] (F = 44.161; p < 0.001), as well as for [E] (F = 20.460; p < 0.001) and [F] (F = 13.839; p < 0.001). Furthermore, an increase in mean values was noted with the transition from category [A1] to [A4]. In the case of [E], the mean increased from EUR 1409.4 to EUR 2194.0, while for [F], it increased from 482.0 km to 719.4 km. Based on the results, it can be concluded that larger TEs are not only characterized by higher levels of organizational resources but also operate within more intensive and complex operational schemes.

Statistically significant differences were also observed for variables [D] (F = 4.039; p = 0.012), [J] (F = 3.900; p = 0.015), [K] (F = 4.137; p = 0.011), and [NI] (F = 4.905; p = 0.005). It should be noted, however, that the magnitude of these effects is significantly smaller than for the previously discussed variables. These results indicate that as the [A] category increases, TEes are more likely to operate in more complex business profiles. Furthermore, these entities are more likely to perceive organizational benefits resulting from personnel changes and to rate technological and market risks higher. However, no significant main effects were found for variables [G], [HI], and [I], indicating that their levels remain relatively stable and are only slightly influenced by variation by factor [A].

Based on the results of the Fisher LSD test, it can be concluded that differences between groups occur between the extremes of the categories. In the cases of [B] and [C], statistical significance was observed for almost all analyzed contrasts. The exception is the rare comparisons between adjacent mean categories. Similar results were observed for [E] and [F], where groups [A3] and [A4] clearly distinguished themselves from [A1] and [A2]. These results suggest that the influence of factor [A] is not random but takes the form of an ordered gradient.

Variable [B] (Table 6) is one of the most strongly differentiating factors in the entire analyzed system. Particularly pronounced effects were observed for [C] (F = 106.160; p < 0.001) and [A] (F = 58.714; p < 0.001), as well as for [E] (F = 22.942; p < 0.001) and [F] (F = 19.499; p < 0.001). Mean values showed a clear upward trend with increasing vehicle numbers. For example, for variable [E], this increase ranged from EUR 1409.4 in category [B1] to EUR 2257.9 in [B4]. In the case of [F], an increase from 482.0 km to 742.5 km was recorded. The results indicate that fleet expansion at a larger organizational scale generates higher labour costs and is characterized by more intensive use of available resources.

Statistically significant effects were also observed for variables [D] (F = 6.946; p = 0.001), [J] (F = 4.381; p = 0.009), [K] (F = 3.330; p = 0.028), [L] (F = 2.831; p = 0.049), [M] (F = 5.428; p = 0.003), and [NI] (F = 5.674; p = 0.002). Furthermore, the lack of significant effects for [G], [HI], and [I] is noteworthy, suggesting that fleet size differentiates “hard” organizational parameters and risk perceptions to a greater extent than general assessments of systemic benefits.

The LSD test results show that contrasts between the smallest and largest fleet sizes were crucial. In the case of variable [C], significance was observed for all comparisons between groups. This indicates a clearly structured relationship. For [E], [F], and [J] ratings and variables related to technological threats, the [B4] category was particularly prominent, with higher mean values than in groups with fewer vehicles. An extensive fleet indicates a larger scale of operations and a greater awareness of technological and competitive threats.

[C] also clearly differentiated the studied TEs (Table 7). The strongest effects were obtained for the number of trucks (F = 91.796; p < 0.001), [A] (F = 43.579; p < 0.001), [E] (F = 26.238; p < 0.001), and [F] (F = 16.402; p < 0.001). As [C] increased, so did [E]—from EUR 1380.0 in [C1] to EUR 2257.9 in [C4]—and [F]—from 505.0 km to 742.5 km. This pattern of results confirms that [C] is a good indicator of overall organizational maturity and the scale of transport operations.

Significant differences were also found for [D] (F = 5.452; p = 0.003), [I] (F = 3.398; p = 0.025), [J] (F = 5.791; p = 0.002), [M] (F = 6.389; p = 0.001), and [NI] (F = 5.048; p = 0.004). In the case of [I], the highest mean was achieved by [C3] (mean = 4.8), suggesting that [A3] may be particularly sensitive to the savings potential of the analyzed solutions. For the benefits associated with [J], the highest means were observed for [C3] and [C4], confirming that, with larger employment scales, the problems of labour costs and driver availability become more acute.

Post hoc analysis indicates that the greatest differences were mainly revealed between the smallest and largest groups. This was particularly true for [E], [F], and [I], as well as risk assessments. It is also noticeable that the selected effects did not increase in a perfectly linear manner but reached a maximum in medium- or large-sized groups, which may reflect a non-linear relationship between the scale of employment and perceptions of effectiveness and risk.

Variable [D] (Table 9) significantly differentiated the basic structural parameters of TEes: [A] (F = 6.524; p = 0.001), [B] (F = 9.602; p < 0.001), and [C] (F = 8.042; p < 0.001). Based on the mean analysis, it can be concluded that the values of these variables are significantly higher in categories that carry out [D3] and [D4] transport than in entities that handle [D1] and [D2]. Therefore, the greater diversity of transport assortment is associated with the larger scale and greater complexity of companies’ operations.

The results of the means and p-values allow us to conclude that in this part of the analysis, differences also occur in selected operational variables and in the structure of transport. The largest differences are primarily observed between categories with low and high levels of diversification. However, the general trend is quite consistent: TEes that handle a larger number of commodity groups have more extensive resources and a more complex business profile. Furthermore, caution should be exercised when interpreting this part of the results. Label shifts were observed in the table, which means that this part of the measurements may be less clear than the results for factors [A], [B], and [C].

Analysis of dichotomous variables regarding the types of goods transported reveals several distinct profiles of TEs (Table 10, Table 11, Table 12 and Table 13). Companies transporting [P1] were, on average, larger and had higher [B] and [C] values compared to companies not serving this market segment. At the same time, it should be noted that they achieved higher [E] (EUR 2229.0 vs. EUR 1681.9) and operated longer routes, as measured by [F] (719.0 vs. 585.1 km). Furthermore, they also rated [M] more favourably. These results suggest that operating [P1] is associated with a more developed and intensive business model.

A similar situation occurs for [P2]. TEs operating this type of transport were, on average, larger considering [A], [B], [C], and [D]. At the same time, they achieved higher [E] (EUR 2052.1 vs. EUR 1679.5) and operated longer routes, as measured by [F] (706.7 vs. 571.3 km). These companies also more frequently indicated a greater importance of [J] and rated [M] and [NI] higher. Based on a comprehensive analysis of the results, it can be concluded that [P2] service is associated not only with more complex operational activities but also with greater risk awareness among companies.

For [P3] and [P4], the differences between the groups were significantly smaller. In the case of [P3], significant differences mainly concerned [D] and the frequency of this type of transport occurring together with other types. However, for [P4], significant differences occurred primarily in the scope of [D]. This also indicates that these two types of activities do not distinguish the studied entities as strongly as [P1] and [P2], especially with respect to organizational scale or basic economic parameters.

[P5] is characterized by a clear but distinct pattern. TEs operating this type of transport were, on average, smaller in terms of [A], [B], and [C]. They also achieved lower [E] (EUR 1431.6 vs. EUR 1993.6) and operated shorter routes, as measured by [F] (549.7 vs. 646.9 km). At the same time, these entities rated [G] higher and [M] and [NI] lower. These results indicate that [P5] operates in the sample on a more limited scale and with a different risk model than TEs operating in more specialized market segments.

For [P6], significant differences were essentially limited to [D]. In the case of [P7], significant differences primarily concerned [F] and [D]. The results indicate that TEs operating [P7] serve longer routes [F] and operate in a more diverse transport environment.

ANOVA analysis shows that the most important factors differentiating the studied TEs include [A], [B], and [C]. These factors most strongly influenced the differentiation of both operational and economic parameters, including [E], [F], the scope of activity, and selected risk assessments. These results indicate that the structural scale of TEs is the key dimension organizing the observed differences in the studied sample.

The second important finding is that larger and more resourceful TEs not only conduct more intensive operations but also perceive [J] more frequently and rate [K], [L], [M], and [NI] higher. This indicates that as TEs develop organizationally, a more complex and simultaneously more realistic perception of both benefits and potential threats emerges.

The third finding concerns [P]. Specialized transport, in particular [P1] and [P2], is associated with larger-scale operations, higher wages, and more intensive transport work. In contrast, [P5] in the studied sample is characterized by a smaller organizational scale, lower [E], and shorter [F]. Furthermore, different perceptions of benefits and threats were observed.

The fourth conclusion relates to the methodological approach. A post hoc Fisher LSD test showed that differences most frequently appeared between extreme categories. This indicates that the phenomena studied are rather gradual in nature and increase with the scale of operations. In practice, no simple division into “different” and “non-different” groups was observed. Rather, an ordered gradient of changes from the smallest entities to the largest emerged.

In summary, the results of the single-factor analysis of variance confirm that the studied TEs differ in an ordered and statistically significant manner, primarily in terms of scale of operations, operational parameters, and some assessments of benefits and threats. This is particularly noticeable in the case of differences resulting from structural factors. However, some variables related to benefit assessment are relatively stable across categories. Therefore, in further analyses, it is worth treating [A], [B], and [C] as key explanatory variables.

3.4. Analysis of Survey Results from the Feeling of Safety When Using Autonomous Vehicles and Heavy Goods Vehicles on the Roads

The study was supplemented by another survey on the sense of safety when using AVs and SAVs on roads. Its primary goal was to assess public awareness and readiness for such a significant leap forward and the introduction of AVs. The 550 people from across Poland participated in the survey. The sample was randomly selected using online communication sources. The results are as follows:

The sample was randomly selected. The surveys were distributed via social media. The study group was 39,82% female, while 60,18% were male (

Table 14. Research results of 550 randomly selected respondents.

		Sex of Respondents [S]	Woman [S1]		Man [S2]
1.	Age of respondents [AR]	18–25 years [AR1]	150	27.27%	100	18.18%
		26–40 years [AR1]	129	23.45%	90	16.36%
		41–55 years [AR]	30	5.45%	20	3.64%
		55 years and over [AR]	22	4.00%	9	1.64%
2.	Driving licence [T]	Yes [T1]	294	53.45%	201	36.55%
		No [T2]	37	6.73%	18	3.27%
3.	Road user safety assessment [U]	1—Very low [U1]	0	0.00%	0	0.00%
		2—Low [U2]	59	10.73%	35	6.36%
		3—Medium [U3]	123	22.36%	77	14.00%
		4—High [U4]	100	18.18%	75	13.64%
		5—Very high [U5]	0	0.00%	0	0.00%
4.	Knowledge of autonomous vehicle operation [W]	Yes [W1]	265	48.18%	192	34.91%
		No [W2]	66	12.00%	27	4.91%
5.	Pedestrian safety perception at pedestrian crossings [Y]	Yes [Y1]	89	16.18%	93	16.91%
		No [Y2]	242	44.00%	126	22.91%

). Women were less interested.

The two largest groups in the study group are aged 18 to 40. This is due to their high interest in new solutions and road safety.

The question regarding driving licences indicates that the majority (90%) hold a driving license. This is directly related to respondents’ ages.

The first four questions aimed to define the study group, gender, age, knowledge of the principles of operation of AVs, and whether they hold a driving licence. The majority of respondents (83%) are aware of the operation of AVs. Based on this group, their opinion on safety as RUs (drivers, pedestrians, and cyclists) was assessed. They were asked whether they would feel safe if AVs were introduced. Their responses were presented as a rating.

As the chart above shows, 36% of survey respondents rated their safety as a 3, and approximately 32% rated it as a 4. These extreme ratings represent a small percentage of the survey group, suggesting that people are open to AVs but not entirely confident in the new solutions.

The next question was whether you would feel safe at a pedestrian crossing knowing there was no driver in the vehicle; 67% of respondents answered no. Again, this may stem from a lack of confidence in new technologies, but it is also important to remember that pedestrian crossings will never be 100% safe.

When asked whether AVs are the future and whether respondents would use them if given the opportunity, responses were split almost 50/50.

Student’s t-test (

Table 15. t-Studenta Test results.

				[S]	[AR]	[T]	[W]	[U]	[Y]
1.	[S]	[S1]	Mean	-	1.770	0.888	0.801	3.118	0.269
			Standard Deviation	-	0.871	0.316	0.400	1.030	0.444
		[S2]	Mean	-	1.717	0.918	0.877	3.256	0.425
			Standard Deviation	-	0.797	0.275	0.330	1.013	0.495
			Cohen’s d	-	−0.06	0.1	0.2	0.13	0.33
			T	-	−0.729	1.132	2.339	1.547	3.845
			Df	-	548	548	548	548	548
			p-value	-	0.466	0.258	0.020	0.123	0.000
			F-ratio Variances	-	1.196	1.314	1.475	1.0349	1.246
			p Variances	-	0.153	0.029	0.002	0.788	0.073
2.	[T]	[T1]	Mean	1.406	1.747	-	0.865	3.293	0.309
			Standard Deviation	0.492	0.822	-	0.342	0.957	0.463
		[T2]	Mean	1.327	1.764	-	0.527	2.091	0.527
			Standard Deviation	0.474	1.018	-	0.504	0.986	0.504
			Cohen’s d	−0.02	0.93	-	1.25	−0.47	0.16
			t	1.132	−0.135	-	6.565	8.809	−3.288
			df	548	548	-	548	548	548
			p-value	0.258	0.893	-	0.000	0.000	0.001
			F-ratio Variances	1.078	1.534	-	2.165	1.062	1.186
			p Variances	0.756	0.022	-	0.000	0.723	0.360
3.	[W]	[W1]	Mean	1.420	1.652	0.937	-	3.407	0.394
			Standard Deviation	0.494	0.746	0.244	-	0.889	0.489
		[W2]	Mean	1.290	2.226	0.720	-	2.022	0.022
			Standard Deviation	0.456	1.095	0.451	-	0.859	0.146
			Cohen’s d	0.27	−0.7	0.75	-	1.57	0.83
			t	2.338	−6.189	6.565	-	13.777	7.271
			df	548	548	548	-	548	548
			p-value	0.020	0.000	0.000	-	0.000	0.000
			F-ratio Variances	1.172	2.154	3.418	-	1.070	11.248
			p Variances	0.353	0.000	0.000	-	0.706	0.000
4.	[Y]	[Y1]	Mean	1.511	1.879	0.841	0.989	3.797	-
			Standard Deviation	0.501	0.877	0.367	0.105	0.826	-
		[Y2]	Mean	1.342	1.685	0.929	0.753	2.864	-
			Standard Deviation	0.475	0.818	0.257	0.432	0.973	-
			Cohen’s d	0.35	0.23	−0.3	0.66	1.01	-
			t	3.845	2.559	−3.288	7.271	11.104	-
			df	548	548	548	548	548	-
			p-value	0.000	0.011	0.001	0.000	0.000	-
			F-ratio Variances	1.113	1.151	2.046	17.079	1.387	-
			p Variances	0.395	0.264	0.000	0.000	0.013	-

) for independent samples was used to determine whether the means of the analyzed variables differed significantly between two independent groups of respondents. The study compared, among others, [S1] and [S2], [T1] and [T2], [W1] and [W2], and [Y1] and [Y2]. Interpretation of the results was based on the t-statistic, the level of statistical significance (p < 0.05), and the effect size, expressed as Cohen’s d. Classic criteria for interpreting effect sizes were adopted: small (d ≈ 0.20), moderate (d ≈ 0.50), and large (d ≈ 0.80).

From a methodological perspective, the selected variables were binary or ordinal and were numerically coded for statistical analyses. Therefore, the mean values calculated for these variables should be interpreted as the mean values of the assigned response codes, not as the classic means for variables measured on a continuous scale. In practice, this does not limit the ability to assess the significance of differences between the compared categories. However, caution is required when interpreting the direction of the relationship when the coding scheme is not clearly defined. The interpretation of variables such as [AR] and [U] remains particularly clear: higher values correspond to higher response categories.

Comparative analysis of categories [S1] and [S2] revealed no statistically significant differences in the analyzed variables. [S] did not significantly differentiate [AR] (t = −0.729; p = 0.466), [T] (t = 1.132; p = 0.258), or [U] (t = 1.547; p = 0.123). These results indicate that respondents from groups [S1] and [S2] were comparable in terms of [AR] structure, [T] frequency, and [U] level.

Statistically significant differences occurred in variables [W] and [Y]. Group [S2] achieved a higher mean value of [W] compared to group [S1] (M = 0.877 vs. 0.801; t = 2.339; p = 0.020), although the observed effect size was small (d ≈ 0.20). An even more pronounced difference was found in the case of [Y], where respondents from group [S2] also achieved a higher mean than those from group [S1] (M = 0.425 vs. 0.269; t = 3.845; p < 0.001). In this case, the effect size was small to moderate (d ≈ 0.33). The results indicate that [S] does not significantly differentiate the basic characteristics of respondents, but is associated with slightly different levels of [W] and [Y].

Considering the division of respondents into [T1] and [T2], no statistically significant differences were found in either [S] (t = 1.132; p = 0.258) or [AR] (t = −0.135; p = 0.893). These results indicate that in the studied sample, [T] was not significantly associated with either of these two characteristics.

It should also be emphasized that individuals from group [T1] reported [W1] significantly more often than those from group [T2] (M = 0.865 vs. 0.527; t = 6.565; p < 0.001). An even more pronounced difference was observed for the [U] variable—the [T1] group achieved significantly higher values (M = 3.293) than the [T2] group (M = 2.091; t = 8.809; p < 0.001). A statistically significant difference was also observed for the [Y] variable (t = −3.288; p = 0.001).

Overall, the results indicate that having a driving licence is associated with a significantly more established understanding of autonomous vehicles and a higher level of [U].

The strongest differentiation occurs in the [W1] and [W2] groups. Individuals in category [W1] were, on average, younger than those in group [W2], as evidenced by the lower mean value of [AR] (M = 1.652 vs. 2.226; t = −6.189; p < 0.001). The effect size was moderate to large (d ≈ −0.70). Furthermore, respondents in group [W1] were more likely to declare [T1] (M = 0.937 vs. 0.720; t = 6.565; p < 0.001) and were also slightly more likely to differ from group [W2] in terms of the structure of [S] (t = 2.338; p = 0.020). However, it should be noted that the effect size was small in this case.

The biggest differences were observed for variables [U] and [Y]. Respondents from group [W1] obtained significantly higher [U] values compared to those from group [W2] (M = 3.407 vs. 2.022; t = 13.777; p < 0.001). In this case, a very large effect size (d ≈ 1.57) was observed. Furthermore, for the variable [Y], a significant and clear difference was noted (M = 0.394 vs. 0.022; t = 7.271; p < 0.001), with a large effect size (d ≈ 0.83). The results suggest that [W] is one of the key factors distinguishing respondents’ attitudes. This is related to both the lower age of [AR], the more frequent occurrence of [T1], and the clearly more positive assessment of [U].

Comparison of groups [Y1] and [Y2] revealed that variable [Y] remained significantly associated with several other respondent characteristics. The results indicated statistically significant differences in [S] (t = 3.845; p < 0.001), [AR] (t = 2.559; p = 0.011), and [T] (t = −3.288; p = 0.001). This suggests that pedestrian safety perception is not an isolated characteristic but rather coexists with a specific socio-demographic profile of respondents.

The strongest differences in this case again concerned variables [W] and [U]. Respondents in category [Y1] achieved a higher mean value of [W] compared to those in group [Y2] (M = 0.989 vs. 0.753; t = 7.271; p < 0.001). The effect size in this case was moderate to large (d ≈ 0.66). Furthermore, an even bigger difference was observed for the [U] variable. Group [Y1] rated it significantly higher (M = 3.797) than group [Y2] (M = 2.864; t = 11.104; p < 0.001). The effect size, in turn, was large (d ≈ 1.01). The results indicate that [Y] is strongly associated with both the overall assessment of road safety and the level of self-reported knowledge of issues related to autonomous vehicles.

The table (Table 14) also presents tests for equality of variances. In some comparisons, the assumption of homogeneity of variances was violated. In particular, in selected comparisons involving [W], [T], and [Y], the significance values for the equality-of-variance tests were below 0.05. It should be noted that the variance in the results across the compared groups was not uniform in these cases. Due to the large sample size, Student’s t-test remains relatively robust to moderate violations of this assumption. Therefore, some results should be interpreted with greater caution or, in cases of significant violations, confirmed using Welch’s t-test, which does not require equal variances.

Based on the analysis, it can be concluded that the strongest differentiating factors among the surveyed respondents were not [S], but primarily the declared [W] and [Y]. The [S] variable differentiated only selected areas, and usually to a small extent. In contrast, variables related to knowledge about autonomous vehicles and safety assessments generated more pronounced and consistent effects across the analyzed comparisons. Secondly, [T1] was associated with higher [U] levels and more frequent [W1] occurrences. Based on this, it can be concluded that practical experience participating in road traffic may foster greater understanding of transportation issues and a more grounded assessment of safety. Thirdly, respondents who declared [W1] form a clearly different profile. They are, on average, younger, more likely to hold a driving licence, and have significantly higher road safety ratings. It is in this category of respondents that the largest statistical effects were observed, particularly with respect to the [U] variable. Fourthly, the [Y] variable shows strong correlations with [U] and [W1]. Therefore, it follows that the perception of pedestrian safety is part of a broader pattern of behaviour toward safety and new transportation technologies. The results of Student’s t-tests indicate that variables related to transportation experience, knowledge of autonomous vehicles, and subjective safety assessments are interrelated. The largest heterogeneity does not occur between [S1] and [S2], but between individuals with different levels of knowledge about autonomous vehicles and between respondents differing in [U] and [Y].

The Kruskal–Wallis test (

Table 16. Kruskal–Wallis Test and Test post hoc results.

No.	Dependent Variable	Grouping Variable	n	Rank Sum	Mean Rank	Mean Response	Standard Deviation	Kruskal–Wallis Test		Test Post Hoc
										Pair of Variables	p-Value	Pair of Variables	p-Value
1.	[S]	[AR1]	250	69,000.0	276.000	0.4	0.491	H	1.655	[AR1] & [AR2]	-	[AR2] & [AR4]	-
		[AR2]	219	61,104.0	279.014	0.411	0.493	N	550	[AR1] & [AR3]	-	[AR3] & [AR4]	-
		[AR3]	50	13,800.0	276.000	0.4	0.495	df	3	[AR1] & [AR4]	-
		[AR4]	31	7621.0	245.839	0.29	0.461	p-value	0.6469	[AR2] & [AR3]	-
2.	[T]	[AR1]	250	67,500.0	270.000	0.88	0.3256	H	7.279	[AR1] & [AR2]	-	[AR2] & [AR4]	-
		[AR2]	219	62,507.0	285.420	0.9361	0.2452	N	550	[AR1] & [AR3]	-	[AR3] & [AR4]	-
		[AR3]	50	13,775.0	275.500	0.9	0.303	df	3	[AR1] & [AR4]	-
		[AR4]	31	7743.0	249.774	0.8065	0.4016	p-value	0.0635	[AR2] & [AR3]	-
3.	[W]	[AR1]	250	72,250.0	289.000	0.880	0.326	H	48.433	[AR1] & [AR2]	1.000	[AR2] & [AR4]	0.001
		[AR2]	219	62,543.0	285.580	0.870	0.339	N	550	[AR1] & [AR3]	0.084	[AR3] & [AR4]	0.688
		[AR3]	50	11,425.0	228.500	0.660	0.479	df	3	[AR1] & [AR4]	0.001
		[AR4]	31	5307.0	171.190	0.450	0.506	p-value	0.000	[AR2] & [AR3]	0.131
4.	[U]	[AR1]	250	62,921.0	251.684	3.02	1.020	H	12.23752	[AR1] & [AR2]	0.0241	[AR2] & [AR4]	1.000
		[AR2]	219	64,384.5	293.993	3.29	1.008	N	550	[AR1] & [AR3]	0.0889	[AR3] & [AR4]	1.000
		[AR3]	50	15,584.0	311.680	3.4	1.010	df	3	[AR1] & [AR4]	1.000
		[AR4]	31	8635.5	278.565	3.19	1.078	p-value	0.0066	[AR2] & [AR3]	1.000
5.	[Y]	[AR1]	250	65,375.0	261.500	0.880	0.326	H	7.279	[AR1] & [AR2]		[AR2] & [AR4]
		[AR2]	219	61,305.5	279.934	0.936	0.245	N	550	[AR1] & [AR3]		[AR3] & [AR4]
		[AR3]	50	15,825.0	316.500	0.900	0.303	df	3	[AR1] & [AR4]
		[AR4]	31	9019.5	290.952	0.807	0.402	p-value	0.064	[AR2] & [AR3]
6.	[S]	[U1]	37	9442.0	255.189	1.324	0.468	H	2.502	[U1] & [U2]	-	[U2] & [U4]	-
		[U2]	94	25,229.0	268.394	1.372	0.483	N	550	[U1] & [U3]	-	[U2] & [U5]	-
		[U3]	200	54,375.0	271.875	1.385	0.487	df	4	[U1] & [U4]	-	[U3] & [U4]	-
		[U4]	175	49,675.0	283.857	1.429	0.495	p-value	0.6442	[U1] & [U5]	-	[U3] & [U5]	-
		[U5]	44	12,804.0	291.000	1.455	0.498			[U2] & [U3]	-	[U4] & [U5]	-
7.	[AR]	[U1]	37	9014.5	243.635	1.595	0.832	H	12.827	[U1] & [U2]	1.000	[U2] & [U4]	0.512
		[U2]	94	24,710.0	262.872	1.702	0.878	N	550	[U1] & [U3]	1.000	[U2] & [U5]	1.000
		[U3]	200	51,579.5	257.898	1.655	0.812	df	4	[U1] & [U4]	0.407	[U3] & [U4]	0.067
		[U4]	175	52,935.5	302.489	1.88	0.782	p-value	0.012	[U1] & [U5]	1.000	[U3] & [U5]	0.960
		[U5]	44	13,285.5	301.943	1.886	0.784			[U2] & [U3]	1.000	[U4] & [U5]	1.000
8.	[T]	[U1]	37	4611.0	124.620	0.351	0.484	H	158.768	[U1] & [U2]	0.000	[U2] & [U4]	1.000
		[U2]	94	27,932.0	297.150	0.979	0.145	N	550	[U1] & [U3]	0.000	[U2] & [U5]	1.000
		[U3]	200	52,625.0	263.130	0.855	0.353	df	4	[U1] & [U4]	0.000	[U3] & [U4]	0.154
		[U4]	175	53,025.0	303.000	1.000	0.000	p-value	0.001	[U1] & [U5]	0.000	[U3] & [U5]	1.000
		[U5]	44	13,332.0	303.000	1.000	0.000			[U2] & [U3]	0.869	[U4] & [U5]	1.000
9.	[W]	[U1]	37	5039	136.1892	0.3243	0.4746	H	196.652	[U1] & [U2]	1.000	[U2] & [U4]	0.000
		[U2]	94	16,793	178.6489	0.4787	0.5022	N	550	[U1] & [U3]	0.000	[U2] & [U5]	0.000
		[U3]	200	61,375	306.875	0.945	0.2286	df	4	[U1] & [U4]	0.000	[U3] & [U4]	1.000
		[U4]	175	54,150	309.4286	0.9543	0.2095	p-value	0.000	[U1] & [U5]	0.000	[U3] & [U5]	1.000
		[U5]	44	14,168	322	1.000	0.000			[U2] & [U3]	0.000	[U4] & [U5]	1.000
10.	[Y]	[U1]	37	7101.5	191.9324	0.027	0.1644	H	109.6855	[U1] & [U2]	1.000	[U2] & [U4]	0.000
		[U2]	94	20,368	216.6809	0.117	0.3232	N	550	[U1] & [U3]	0.556	[U2] & [U5]	0.000
		[U3]	200	49,275	246.375	0.225	0.4186	df	4	[U1] & [U4]	0.000	[U3] & [U4]	0.000
		[U4]	175	57,587.5	329.0714	0.5257	0.5008	p-value	0.000	[U1] & [U5]	0.000	[U3] & [U5]	0.000
		[U5]	44	17,193	390,75	0,75	0,438			[U2] & [U3]	1.000	[U4] & [U5]	0.214

) was used to determine whether the distributions of analyzed variables differed significantly between more than two independent groups. This method is a nonparametric equivalent of one-way analysis of variance. It is particularly useful for ordinal or binary data, or when the assumptions of normality are not met. The analysis primarily assessed differences related to the variables [AR] and [U], which served as grouping variables. [S], [T], [W], [U], and [Y] were included as dependent variables.

Results were interpreted using the H statistic, the degrees of freedom, and a significance level of p < 0.05. If a significant overall result was obtained, a post hoc analysis was additionally conducted to determine which groups of variables differed. Mean ranks, indicating the direction and relative strength of the observed differences between the compared groups, were also an important element of the interpretation.

The [AR] variable did not significantly differ across [S] levels. Similarly, no significant differences were observed in the [T] range. However, the obtained result was close to the assumed level of statistical significance (H = 7.279; p = 0.0635).

A significant effect of [AR] was observed for the [W] variable. The test statistic was H = 48.433, p < 0.001, indicating a clear difference between the analyzed age categories. The highest mean ranks were obtained in the [AR1] (MR = 289.0) and [AR2] (MR = 285.58) groups. The lowest values were recorded in the [AR4] group (MR = 171.19). Post hoc analysis revealed significant differences between the [AR1] and [AR4] groups (p = 0.001) and [AR2] and [AR4] groups (p = 0.001). The remaining comparisons did not reach statistical significance. These results suggest that the level of [W] was significantly lower in the oldest group of respondents than among the younger ones.

The [AR] variable also significantly differentiated [U] levels. For this variable, a value of H = 12.23752 was obtained, at p = 0.0066. This result indicates differences between the analyzed groups. The lowest mean value was recorded in group [AR1] (MR = 251.684). However, higher values were obtained in groups [AR2] (MR = 293.993) and [AR3] (MR = 311.680). According to the post hoc analysis, one statistically significant difference was observed between [AR1] and [AR2] (p = 0.0241). Accordingly, it can be concluded that respondents from group [AR2] rated [U] higher than those in the youngest age category. Despite the highest mean rank being recorded in the [AR3] group, the differences between this group and the others did not reach statistical significance. However, no significant effect of the [AR] variable on the [Y] level was found.

The [U] variable did not significantly differ across [S] levels. A significant result was obtained for [AR], analyzed as a dependent variable relative to the [U] level. For this relationship, the value was H = 12.827 at p = 0.012. This indicates differences between the compared groups. Mean ranks were significantly higher in [U4] (MR = 302.489) and [U5] (MR = 301.943) than in [U1], [U2], and [U3]. Furthermore, post hoc analysis did not reveal unequivocally significant differences between individual pairs of groups, although the comparison between [U3] and [U4] approached statistical significance (p = 0.067). Therefore, these results indicate the existence of a general relationship, but it does not take the form of strong and clearly demarcated contrasts between individual categories. Therefore, a moderate tendency can be observed, according to which a higher level of [U] coincides with a slightly higher [AR].

A very strong effect was observed for the [T] variable. The test statistic was H = 158.768 at p = 0.001. The lowest mean value was observed in the [U1] group (MR = 124.620). However, significantly higher values were observed in the [U2], [U3], [U4], and [U5] categories, particularly in the latter two categories (both MR = 303.0). Furthermore, post hoc analysis revealed significant differences between the [U1] group and all other groups, while no significant differences were observed between the [U2], [U3], [U4], and [U5] categories. These results allow us to conclude that respondents who rated [U] very low clearly differ from all other respondents in terms of [T]. It should be noted that at higher [U] levels, this relationship stabilizes.

An even stronger and more transparent pattern of relationships was observed for the [W] variable. The Kruskal–Wallis test yielded H = 196.652, p < 0.001. Mean values increased almost monotonically from groups [U1] (MR = 136.1892) and [U2] (MR = 178.6489), through [U3] (MR = 306.875) and [U4] (MR = 309.4286), to the highest value in group [U5] (MR = 322.0). Post hoc analysis revealed no significant differences between groups [U1] and [U2], or between categories [U3], [U4], and [U5]. However, significant differences occurred between groups with low [U] levels and groups with medium and high levels. This was true for comparisons of [U1] with [U3], [U4], and [U5], and [U2] with [U3], [U4], and [U5] (all p = 0.000). These results indicate that higher levels of [U] are strongly associated with higher levels of [W1].

A very clear effect was also noted in the case of the [Y] variable. The Kruskal–Wallis test yielded H = 109.6855 (p < 0.001), indicating significant differences between groups distinguished by [U] level. The mean ranks increased with the increase in the road user safety assessment—from MR = 191.9324 in group [U1], through 216.6809 in [U2] and 246.375 in [U3], to 329.0714 in [U4] and 390.75 in [U5]. Post hoc analysis showed that significant differences occurred primarily between groups with low and medium [U] levels and groups with high and very high assessments. This applied to comparisons of [U1] with [U4] and [U5], [U2] with [U4] and [U5], and [U3] with [U4] and [U5] (in all cases p = 0.000).

However, no significant differences were found between groups [U1] and [U2], [U1] and [U3], [U2] and [U3], or between groups [U4] and [U5]. This result indicates a clear division of respondents into two clusters: those with low and average ratings of level [U] and those with high and very high ratings. Respondents belonging to the latter group also had significantly more positive perceptions of level [Y].

The analysis shows that the [AR] variable significantly differentiates the levels of [W] primarily and [U1]. However, it does not significantly affect [S], [T], or [Y]. The most pronounced relationship is observed for the [W] variable, whose level was significantly lower among respondents in the oldest group than among those in the younger groups. Secondly, the fundamental importance of the [U] variable should be emphasized. It did not significantly differentiate [S] levels but remained strongly associated with [AR], [T], [W], and [Y]. Particularly pronounced effects were observed for [T] and the variable on knowledge of autonomous vehicles, where individuals with very low ratings of [U] clearly differed from other respondents on the analyzed characteristics. Third, based on the obtained results, a significantly consistent pattern of attitudes related to safety can be identified. Respondents with higher ratings of [U] more often declare [W1] and have a more positive assessment of pedestrian safety at crosswalks. This means that [U] is not an isolated characteristic in the studied sample, but remains associated with a broader, more positive attitude toward safety and transportation technologies. Fourth, based on the post hoc analysis, it can be concluded that the differences are not entirely linear but often form distinct “blocks” of categories. In particular, in the case of knowledge of autonomous vehicles and [Y], it is possible to distinguish categories of respondents with lower [U] values and groups with higher [U] values, which differ more markedly from each other than from their immediate neighbouring categories.

In summary, the results of the Kruskal–Wallis test and post hoc analyses indicate that in the analyzed sample, the most significant relationships concern the variables [AR], [T], [W], and [U]. The strongest and most unambiguous effects were observed in those areas where the relationships between the general assessment of [U] and more specific indicators of attitudes and transportation competencies were analyzed.

3.5. Assessment of the Impact of Autonomous and Semi-Autonomous Vehicles on Freight Transport

For clarity in the analysis and its development, the five most important factors were selected and described.

Strengths:

S1: Reducing the number of drivers in a company means that fewer drivers will be needed to perform the same tasks and routes, thus reducing costs.

S2: Reducing salaries and fewer drivers means greater savings that can be allocated to investments and development.

S3: Reducing the use of non- AVs (conventional) means fewer jobs and, once again, opportunities for development and investment.

S4: Reducing CO₂ emissions—AVs can reduce CO₂ emissions because their driving is much smoother and they can plan appropriate maneuvers in advance.

S5: Lower maintenance costs translate into savings. These also result from much smoother driving and the planning and anticipation of maneuvers. An AV connected to a telematics system will be informed in advance of a traffic light turning red, allowing it to initiate engine braking quickly without using the foot brake.

Weaknesses:

W1: High purchase cost of an AV. Sensors and cameras equipped with an AV will increase the purchase price.

W2: Poorly adapted infrastructure. The infrastructure for AV driving is poor, and its development is both costly and time-consuming. Furthermore, its development may vary across countries.

W3: Loss of driver jobs means the loss of good and trusted employees who will become redundant.

W4: Poorly developed regulations. The regulations governing AVs are not fully defined, and there are issues with criminal liability and insurance coverage in the event of an accident or collision.

W5: Deterioration in the quality of transport services. The introduction of AVs will lead to a lack of direct contact, as the machines will not be able to answer questions.

Opportunities:

O1: Acquiring new orders. By using AVs, it will be possible to reduce the price of the service and order provided and increase market competitiveness.

O2: Covering longer routes. An AV can drive without interruption and cover longer distances in a shorter time than conventional vehicles.

O3: Access to new technologies. The company’s development and the purchase of new vehicles will provide access to previously unused technologies that will help improve the transport and monitoring of goods.

O4: Development of the transport industry. This will occur with the increase in the volume of transported cargo.

O4: Expanding the fleet; the savings can be allocated to the development and purchase of new AVs.

Threats T:

T1: Loss of communication with the vehicle. A very high risk that could cause numerous problems.

T2: Risk of system failure. A system failure could cause significant problems, including significant delays in transport. Repairing the system and troubleshooting the problem are time-consuming.

T3: Risk of sensor damage. Mechanical damage to the sensors, for example, during unloading, could result in the vehicle becoming immobilized and unable to continue operating.

T4: Risk of accident. Dirty radars or cameras, or light reflections, could lead to an accident.

T5: Possibility of competitors emerging with better solutions.

3.6. SWOT/TOWS Analysis of Autonomous Vehicle Adoption in Transport Enterprises

The highest number of interactions (64/2) and the highest weighted number of interactions (12.35) in the analyzed case occur between strong sides and opportunities. This means that companies introducing new solutions to the market, such as AVs and SACs, should adopt an aggressive strategy. The highest number of interactions (64/2) and the highest weighted number of interactions (12.35) in the analyzed case occur between strong sides and opportunities. This means that companies introducing new solutions to the market, such as AVs and SACs, should adopt an aggressive strategy (

Table 17. Summary of SWOT and TOWS analyses.

Combinations	SWOT Analysis Results		TOWS Analysis Results		Summary Table
	Sum of Interactions	Sum of Products	Sum of Interactions	Sum of Products	Sum of Interactions	Sum of Products
Strengths/Opportunities	32/2	6.25	32/2	6.1	64/2	12.35
Strengths/Threats	8/2	1.4	18/2	2.3	26/2	3.7
Weaknesses/Opportunities	28/2	6.1	18/2	3.55	46/2	9.65
Weaknesses/Threats	26/2	3.2	16/2	3.7	42/2	6.9

and

Table 18. Strategy matrix.

	Opportunities	Threats
Strengths	Aggressive Strategy Number of interactions—64/2; Weighted number of interactions 12.35	Conservative Strategy Number of interactions—26/2; Weighted number of interactions 3.7
Weaknesses	Construction Strategy Number of interactions—46/2; Weighted number of interactions 9.65	Defensive Strategy Number of interactions—42/2; Weighted number of interactions 6.9

).

An aggressive strategy (maxi-maxi) suggests that strengths predominate, while opportunities emerge in the environment. Opportunities emerging in the environment will allow for the confident implementation of the plan, and that strengths will help achieve them. The strategy encourages action and the capture of opportunities and possibilities in the environment.

4. Cost Minimization Model and Comparative Results for Conventional and Semi-Autonomous Trucks

4.1. Development of the Cost Minimization Model Within a Sustainability Framework

Improving the sustainability of RFT requires not only technological innovation but also a clear understanding of the associated economic implications. While AV and SAV technologies are frequently discussed in the context of safety and efficiency, their adoption ultimately depends on operational feasibility and cost performance. rom a sustainability perspective, transport solutions must be evaluated through an integrated lens that considers environmental, economic, and resource-efficiency dimensions. In freight TSs, cost per tonne-kilometre represents one of the key indicators linking operational efficiency with economic sustainability. Reductions in this metric may reflect improved fuel efficiency, better asset utilization, and potentially lower environmental impact. To assess the economic and sustainability-related effects of ATes, a cost minimization model was developed. The model aims to quantify the cost structure of freight operations under conventional and semi-autonomous driving conditions, while accounting for differences in fuel consumption, maintenance requirements, labour costs, and IR. The objective function minimizes the cost per tonne-kilometre. For a conventional truck, this indicator can be calculated as follows:

$$C_0={\displaystyle \frac{C_{fuel}\times L+C_{maint}\times L+C_{driver}+A}{Q\times L}}\to min$$

(1)

where $C_0$—Cost per tonne-kilometre of a conventional truck (monetary units).

$C_{fuel}$—Fuel cost per kilometre (fuel consumption in L/km × price per litre), monetary units.

$C_{maint}$—Maintenance and service costs per kilometre, monetary units.

$C_{driver}$—Daily driver labour cost, monetary units.

$A$—Daily depreciation and insurance costs of the truck, monetary units.

$Q$—Payload capacity of the vehicle, expressed in tonnes.

$L$—Daily mileage of the truck, kilometres.

Equation (1) can be alternatively expressed as:

$$\begin{gathered} C_0={\displaystyle \frac{L\times \left(C_{fuel}+C_{maint}\right)+C_{driver}+A}{Q\times L}}= \\ {\displaystyle \frac{C_{fuel}+C_{maint}}{Q}}+{\displaystyle \frac{C_{driver}+A}{Q\times L}}\to min, \end{gathered}$$

(2)

To calculate this indicator for a semi-autonomous truck, it is necessary to introduce a correction coefficient reflecting the level of road IR for autonomous driving. This coefficient is defined as the IRS. The IRS represents the proportion of a given route that can be operated in autonomous mode, while the remaining share (1 − IRS) corresponds to conventional, human-controlled driving conditions. The value of the IRS is determined as a weighted average of individual route segments, taking into account the category and characteristics of the road infrastructure.

–

IRS = 0.9–1.0—for motorways with full traffic-flow separation, absence of at-grade intersections, availability of 5G network coverage, high-quality road markings, and the implementation of Vehicle-to-Infrastructure (V2I) communication technologies. Under such conditions, the highest levels of autonomous driving safety can be achieved, together with the possibility of full platooning (truck convoy operations);

–

IRS = 0.7–0.8—for high-speed highways with good pavement quality, the presence of a median barrier, and a limited number of interchanges. Such road segments demonstrate a high degree of readiness for autonomous driving and platooning operations. However, the occurrence of complex junctions may still require driver intervention;

–

IRS = 0.3–0.4—for interregional roads consisting of two traffic lanes, lacking physical separation, and characterized by the presence of at-grade intersections and pedestrian crossings. On such road segments, the probability of transferring vehicle control back to the HD remains high;

–

IRS = 0.0–0.1—for urban roads with dense traffic conditions, a high degree of unpredictability in the driving environment, and often inadequate or inconsistent road markings. Under such circumstances, the feasibility of autonomous driving is considered to be extremely limited.

In this case, the IRS does not constitute an element of scientific novelty, as similar composite indices have been widely applied in both research and practice in the fields of logistics and autonomous transport in EU countries. In particular, such indices include the Autonomous Vehicles Readiness Index (AVRI), Infrastructure Support for Automated Driving (ISAD), and the Smart Road Infrastructure Classification Index (SRICI), among others.

The platooning effect described above represents one of the most significant factors contributing to fuel savings during autonomous convoy operations. This effect is achieved by maintaining a minimal distance between trucks (typically 10–15 metres), resulting from the following mechanisms:

–

Aerodynamic benefits—the leading vehicle reduces air resistance, while the following trucks operate within a zone of lower pressure;

–

Motion synchronization—V2I communication systems enable coordinated acceleration and braking, thereby mitigating the “accordion effect” and preventing unnecessary fuel consumption.

Empirical studies indicate that fuel savings under platooning conditions may reach approximately 8–11%. Accordingly, the fuel cost per kilometre in semi-autonomous operations can be expressed as a reduced proportion of the fuel cost associated with conventional truck driving.

Maintenance and servicing costs for a SAV, compared to a conventional truck, increase by up to 15% and can therefore be expressed as: ${1.15C}_{maint}$.

In addition, the acquisition cost of a SAV may be up to 30% higher than that of a conventional truck. This difference directly affects depreciation and insurance costs, which can therefore be expressed as an increased proportion $1.3A$ of the baseline level.

Considering the above assumptions and based on the reformulated equation, the objective function minimizing the cost per tonne-kilometre for a semi-autonomous truck can be expressed as follows:

$$C_1={\displaystyle \frac{0.9C_{fuel}\times IRS+C_{fuel}\times \left(1-IRS\right)+1.15C_{maint}}{Q}}+{\displaystyle \frac{C_{driver}+1.3A}{Q\times \frac{L}{\left(1-IRS\right)}}}\to min,$$

(3)

4.2. Input Parameters and Model Assumptions

The practical applicability of the proposed cost minimization models depends on the appropriate specification of input parameters reflecting real operating conditions in RFT.

The input parameters used in Models (2) and (3) were determined based on data obtained from the [294,295] are presented in

Table 19. Input parameters for conventional and semi-autonomous truck operations.

Parameter	Conventional Truck	Semi-Autonomous Truck
Fuel cost per kilometre, Cfuel (EUR);	0.4703	0.4232
Maintenance and service cost per kilometre, Cmaint (EUR);	0.12	0.138
Daily driver labour cost, Cdriver (EUR);	150	150
Daily depreciation and insurance cost, A (EUR);	85	110.5
Payload capacity, Q (tonnes);	22	22
Normative daily mileage, L (km)	650	1400

. For the calculation of fuel cost per kilometre, the average fuel consumption was assumed to be 28.5 L per 100 km, while the diesel fuel price was set at EUR 1.65 per litre.

Models (2) and (3) enable a comparative analysis of unit freight transport costs under conventional and semi-autonomous operating conditions, taking into account the weighted IRS calculated for a specific truck-platooning route. The results of this comparative analysis are presented in

Table 20. Comparison of cost per tonne-kilometre for conventional and semi-autonomous trucks across different Infrastructure Readiness Score (IRS) levels.

IRS for Autonomous Driving	Cost Per Tonne-Kilometre for a Conventional Truck, C0 (EUR)	Maximum Daily Mileage of a Semi-Autonomous Truck, L (km)	Cost Per Tonne-Kilometre for a Semi-Autonomous Truck, C1 (EUR)
0.0	0.0433	650.0	0.0459
0.1	0.0433	722.2	0.0438
0.125 *	0.0433	742.9	0.0433
0.2	0.0433	812.5	0.0418
0.3	0.0433	928.6	0.0398
0.4	0.0433	1083.3	0.0377
0.5	0.0433	1300.0	0.0357
0.6	0.0433	1400.0	0.0348
0.7	0.0433	1400.0	0.0346
0.8	0.0433	1400.0	0.0344
0.9	0.0433	1400.0	0.0342
1.0	0.0433	1400.0	0.0340

* The level at which the cost of autonomous driving is equal to driving with a driver.

.

As shown in Table 20, the daily mileage of a semi-autonomous truck gradually approaches its maximum attainable value as the IRS increases up to approximately 0.54. along a given route. Further improvements in road infrastructure do not result in additional mileage growth, as the vehicle reaches its operational limit under the most favourable conditions.

The mathematical justification for this value is based on the constraint that the daily mileage of a truck with a driver does not exceed 650 km per day. In semi-autonomous mode, the truck can travel min(650/(1 − IRS); 1400) km per day.

Accordingly, at IRS = 0.54, the daily mileage in semi-autonomous operation reaches its practical maximum of 1400 km per day.

Consequently, within this IRS range, the cost per tonne-kilometre for semi-autonomous truck operations exhibits a pronounced decrease, as illustrated in Figure 7.

The intersection point of indicators C₀ and C₁ across different IRS values (Figure 7) provides a basis for assessing the economic feasibility of deploying conventional versus semi-autonomous freight vehicles on specific transport routes. In particular, when the average level of IR along a route is below IRS < 0.125, conventional truck operations demonstrate lower unit transport costs. Conversely, for IRS > 0.125, SAVs become the more economically advantageous solution. Within this range, the unit cost C₁ decreases from EUR 0.0433 to EUR 0.0348 per tonne-kilometre.

The indicated value of IRS = 0.125 was obtained numerically from the identity C₀ = C₁, i.e., when the cost per ton-kilometre for a conventional truck equals the cost per ton-kilometre for a semi-autonomous truck.

In developing the model, the methodology of scientific inquiry was applied, specifically the deliberate use of the principle of scientific abstraction. Any economic–mathematical model is, by nature, an abstraction—a simplified representation of real-world economic objects or processes expressed through mathematical equations.

It should be emphasized that the purpose of this model was not to analyze macro-level trends in the freight transport industry, the complexity of which is characterized by an infinite number of interconnections and deterministic chaos, but rather to justify the feasibility of implementing semi-autonomous trucks within the operations of an individual logistics company.

Thus, the model has its limitations and represents an isolated system. The example considered was based on specific values of vehicle costs, additional expenses related to equipping trucks with semi-autonomous driving systems, and the level of road infrastructure along the routes of their direct use. The concept of fixed and variable costs is widely used in the economic justification of business decisions, and despite its level of abstraction, it was considered appropriate for application in the construction of this model.

The developed model may therefore support further applications:

–

First, in the presence of SAVs, freight routes may be optimized not only according to the shortest-distance criterion but also with respect to minimizing total transport costs, accounting for variations in IR across route segments;

–

Second, for TEes operating a stable portfolio of freight routes, the model may serve as a tool for economically justifying fleet composition decisions. These directions represent promising avenues for further research.

5. Discussion

Based on the research results, several important points can be drawn regarding the current state and near future of autonomous freight transport systems. First, it should be noted that semi-autonomous vehicles are currently a transitional stage in the process of transport automation. According to the literature, full vehicle autonomy is associated with numerous barriers, including infrastructure requirements, regulatory uncertainty, and ensuring system reliability. However, increasingly common semi-autonomous solutions can be implemented within existing transport infrastructure without requiring organizational or technical changes.

Second, the research results indicate a discrepancy between the expected benefits of automation and the risks perceived by transport companies. Advantages of semi-autonomous and autonomous vehicles include reduced operating costs and increased efficiency. It should also be noted that transport companies express significant concerns about system reliability, connectivity quality, and the level of preparedness of the infrastructure. These conclusions are consistent with previous studies, which indicate that the mere availability of technology does not guarantee its effective implementation and use. Organizational and regulatory factors, as well as the level of trust in new solutions, are also crucial. Third, according to the developed cost minimization model, infrastructure conditions are crucial for assessing the economic viability of semi-autonomous transport. Based on the Infrastructure Readiness Index (IRS), it can be concluded that the benefits of automation vary depending on the specific nature of a given route. Therefore, implementing automated freight transport systems should be a holistic process. In particular, it should take into account the condition of the infrastructure, the regulatory environment, and actual operating conditions.

The obtained results are largely consistent with previous research, which indicates the potential of autonomous and semi-autonomous vehicles in increasing transport efficiency and reducing operating costs. In particular, the observed decrease in cost per tonne-kilometre with increasing infrastructure readiness aligns with literature findings emphasizing the importance of better fleet utilization and smoother and more predictable driving patterns.

Furthermore, the obtained research findings provide a novel insight into economic profitability, which is not an automatic effect of automation implementation. It depends largely on the quality and preparedness of infrastructure. Therefore, it should be stated that cost benefits are conditional, not universal.

In this context, the introduction of the Infrastructure Readiness Index (IRS) allows for a better understanding of the actual operating conditions of the transport system. This approach allows for a more precise determination of the conditions under which semi-autonomous solutions deliver real benefits and where their effectiveness remains limited.

The key contribution of this article to contemporary science is the determination of the break-even point (IRS ≈ 0.125). In practice, this represents a specific effect that indicates real economic benefits. This is significant because many previous studies focused on overall efficiency improvements, without specifying when exactly this would actually pay off.

The results obtained from transport companies largely confirm previous observations that risk, both technological and regulatory, remains the main obstacle to implementation. At the same time, the study shows that this risk is assessed significantly higher compared to the potential benefits.

Similar conclusions emerge from a survey of road users. The study confirms that overall acceptance of such solutions is rather moderate and the level of trust remains limited, especially in safety-related situations. In practice, a key factor in implementing semi-autonomous freight transport technologies is taking actions that genuinely build user confidence.

Based on the research results, the transition to autonomous freight transport does not depend solely on technological development. In practice, it is the result of the interaction of several key factors, including the level of infrastructure preparedness, economic viability, the organization’s ability to implement them, and the degree of social acceptance. This approach broadens the existing view in the literature, indicating that only the combined consideration of the indicated factors allows for a better understanding of the actual conditions and pace of implementation of semi-autonomous transport.

From a sustainability perspective, research results show that semi-autonomous and autonomous vehicles can reduce energy consumption and emissions per ton-kilometre. However, it should be emphasized that these effects depend on the degree to which automation is supported by infrastructure and operational optimization.

Ultimately, it should be concluded that the implementation and use of automated technology in freight transport is determined not only by technological factors, but also by economic, organizational, and institutional factors.

The research findings indicate that the practical application of fully automated vehicles in freight transport requires attention to numerous infrastructural, legal, and technological constraints. Furthermore, the presented research findings fill a research gap, indicating that the economic viability of semi-autonomous solutions is largely dependent on the level of infrastructure readiness. This means that automation should not be viewed solely as a technological innovation, but as a complex, systemic process shaped by many interrelated factors.

Furthermore, previously published studies have shown that automation is directly linked to increased efficiency and improved sustainability. In contrast, the results of this study suggest that these effects are significantly dependent on operational and infrastructural factors. Therefore, this research provides a more advanced perspective on automation and indicates that it is a solution determined by many different circumstances.

6. Conclusions

RFT remains essential for economic activity and supply-chain continuity, yet its long-term development is increasingly constrained by environmental impacts, capacity limits, and a shortage of professional drivers. In this study, the sustainability implications of autonomous and semi-autonomous freight vehicles were examined using a combined empirical–analytical approach, including surveys, SWOT/TOWS assessment, and a cost minimization model that explicitly incorporates IR through the IRS. The modelling results show that the economic and operational performance of semi-autonomous trucking is highly route-dependent. As the share of infrastructure suitable for automated operation increases, semi-autonomous trucks can extend daily mileage and reduce cost per tonne-kilometre, primarily through higher asset utilization and improved fuel efficiency (including platooning-related savings). A clear break-even threshold is observed at IRS ≈ 0.125: when IRS < 0.125, conventional operations remain cost-advantageous; when IRS > 0.125, semi-autonomous operations become economically preferable. The strongest cost improvements occur as IRS rises toward approximately 0.6, after which mileage approaches an operational ceiling and additional infrastructure improvements provide diminishing returns in utilization. From a sustainability perspective, these results are important because unit-cost reductions in the model are coupled with mechanisms that typically lower energy use and emissions intensity per tonne-kilometre, such as smoother driving, reduced congestion effects via more stable traffic behaviour, and improved logistics efficiency through higher vehicle productivity. However, the empirical evidence indicates that current adoption readiness is limited. TEes rate risks—connectivity loss, system failures, sensor damage, and infrastructure gaps—higher than expected benefits, while public opinion remains cautious toward fully AVs. These findings underline that sustainability gains will not materialize through technology alone; they require coordinated progress in infrastructure, regulation, liability/insurance frameworks, and trust-building measures. Overall, the study supports semi-autonomous freight vehicles as the most realistic near-term pathway toward sustainable automation. In practice, the IRS-based model can be used (i) to compare route alternatives not only by distance but also by expected cost and sustainability performance under different infrastructure conditions, and (ii) to justify fleet-structure decisions for companies operating stable portfolios of freight routes. Future research should expand the modelling framework to include explicit emission factors, reliability and downtime costs, and scenario-based policy analysis across cross-border routes with heterogeneous IR.

Based on the research results, it can be concluded that the introduction of semi-autonomous freight transport involves not only the use of appropriate technology but also the implementation of appropriate legal regulations and the efficient management of the entire implementation process. The first stage should focus on preparing the necessary infrastructure, with particular emphasis on road expansion, the development of digital connectivity (e.g., 5G networks), and V2I communication systems. It is important to emphasize that these elements largely determine the level of infrastructure readiness (IRS) and the reliable operation of semi-autonomous vehicles in industry. The second stage involves streamlining and standardizing legal regulations. In this case, basic safety principles, vehicle approval procedures, and liability issues should be defined. It is important to emphasize that transparent rules governing the use of autonomous vehicles by drivers reduce legal risk and can influence business investment decisions. The third stage involves integrating semi-autonomous vehicles into transport companies’ existing fleets. In this regard, companies should gradually introduce semi-autonomous vehicles into their fleets, plan routes, and adapt logistics processes to new conditions. The fourth stage involves widespread adoption both in the freight transport market and across society. Above all, attention should be paid to the level of acceptance of such solutions among the public, changes in the employment structure, and the emergence of new business models in freight transport. Such a comprehensive approach, which considers various levels of management, ensures that the implemented technological solutions will translate into tangible economic and sustainable development outcomes.

Research findings suggest that implementing semi-autonomous freight transport requires coordinated action at multiple levels. Infrastructure development, particularly in digital connectivity and road quality, is crucial for freight transport automation. It should also be emphasized that streamlining and standardizing regulations in this area is essential, as this can reduce uncertainty surrounding liability and system certification. Furthermore, the effective implementation of automation technologies requires organizational and digital changes. A lack of coherent action in these areas may prevent the full realization of automation’s potential benefits.

This study has several limitations that should be considered when interpreting the results. First, the survey of transport companies was conducted with a relatively small sample. Therefore, the obtained results are only preliminary and should not be considered fully representative. Second, the public opinion survey was based on an online, non-probabilistic sample, which may introduce selection bias. Third, the SWOT/TOWS analysis provides structured strategic knowledge but does not allow for statistical verification.

Furthermore, the cost minimization model was developed under simplified assumptions regarding fuel consumption, maintenance costs, and infrastructure availability. It is important to emphasize that this model does not account for stochastic variability, operational disruptions, or detailed emissions modelling. Another significant limitation is that the described model does not address cross-border regulatory differences and the varying legal frameworks within the EU.

Future research should focus on several important aspects. First, the study sample should be enlarged to be more representative and confirm the observed relationships. Second, analytical models should additionally incorporate emission and reliability indicators, as well as uncertainty analysis. Third, research requires attention to the diverse cross-border regulations and their diverse interpretations. Furthermore, future research should include analysis of organizational and digital transformation processes within transport companies, as these can significantly support or constrain the implementation of automation technologies.