How to Facilitate Safety Performance of Human–Robot Collaboration in High-Rise Construction Scenarios: An Empirical Model

Abstract

Despite the growing use of collaborative robots in high-rise construction, ensuring safe human–robot collaboration (HRC) in hazardous environments remains a critical challenge. Addressing the gap that previous studies optimized human, robot, or task factors in isolation without a systemic coordination perspective, this study develops and empirically validates a Human–Task–Robot Alignment (HTRA) framework to explain how alignment mechanisms enhance safety performance in the construction of HRC. Data from 84 high-rise HRC projects were analyzed using Partial Least Squares Structural Equation Modeling (PLS-SEM). The results reveal that HTRA serves as a pivotal mechanism for safety improvement, where Human–Robot Alignment (HRA) and Task–Human Alignment (THA) exert stronger effects on bilateral safety perception than Task–Robot Alignment (TRA), underscoring the centrality of human factors at the current stage of HRC development. Moreover, this study identifies a sequential cognitive–behavioral path from safety perception to safety intention and safety behavior performance, explaining how alignment enhances safety performance. Standardized HRC guidelines further strengthen HRA, facilitating safer and more efficient collaboration. This study extends Task–Technology Fit theory to a triadic human–task–robot context and advances the concept of bilateral safety, providing theoretical and managerial guidance for developing next-generation safe collaboration systems in construction.

1. Introduction

High-rise construction and maintenance are among the most recurrent yet hazardous operations in the built environment, with statistics showing that falls from height account for over half of all construction accidents [1]. Robotics and automation technologies provide promising solutions by replacing human workers in hazardous environments. Typical applications include curtain-wall-cleaning robots developed by Nihon Bisoh Co., Ltd. (Tokyo, Japan) [2,3], façade spray-painting robots introduced by BOZHILIN Co., Ltd. in China [4], and glass curtain wall installation robots [5,6]. Empirical evidence has demonstrated that human–robot collaboration (HRC) in accomplishing construction tasks can significantly reduce labor demand and improve operational efficiency [6,7]. Moreover, it offers additional benefits by promoting sustainability [7,8] and ensuring quality control [5].

However, the integration of collaborative robots (cobots) into high-rise construction introduces new and complex safety challenges that current practices and research have not adequately addressed [9,10]. Unlike factory environments, construction sites are characterized by unstructured, constantly changing conditions, where factors such as wind load, glare, surface irregularity, and narrow operation spaces may destabilize robotic trajectories or impair sensor accuracy. Robots operating at height, such as unmanned aerial systems (UASs) or façade robots, must interact closely with human workers in these highly unstructured environments, which introduces substantial risks of collision, trajectory deviation, or loss of control. Meanwhile, workers often experience psychological stress, distraction, and reduced situational awareness when collaborating with autonomous machines [11,12]. These challenges highlight an urgent need to understand and manage the safety mechanisms underpinning HRC at height, where human and robotic agents must coordinate closely under dynamic, unpredictable conditions. Ensuring this safety is not only vital for protecting workers but also for building trust and accelerating the adoption of HRC in the construction industry.

Existing research on improving HRC safety can be categorized into three dimensions. First, most studies focus on enhancing robotic autonomy and reliability through mechanical design and control optimization. For example, curtain-wall-cleaning robots have been equipped with advanced sensor systems to overcome obstacles [13], and integrated control systems with dual-redundant communication networks to ensure stable execution [7]. In addition, cobots are designed to interpret workers’ intentions (e.g., using posture recognition, skeletal tracking, and electroencephalography (EEG)), and dynamically adjust their behaviors to increase collaboration safety [14]. Second, at the human level, research emphasizes equipping workers with the necessary collaborative knowledge and skills needed through immersive virtual environments, as well as enhancing their safety awareness and situational awareness to prevent accidents [15,16]. Finally, task-based safety improvements focus on adaptive task allocation and scheduling strategies, aiming to minimize spatiotemporal overlap in shared workspaces, thereby reducing collision risks [17].

Although these studies have advanced HRC safety from different perspectives, three key limitations remain. (1) They typically optimize single dimensions in isolation, overlooking the interdependence between humans, robots, and tasks. (2) They often assume stable environments and predictable interactions, which seldom hold true in dynamic, high-rise construction contexts. (3) They lack an integrative theoretical framework to explain how coordination—or misalignment—among these dimensions influences safety outcomes. For example, even advanced robotic technologies can undermine workers’ control if they exhibit excessive autonomy and insufficient transparency, which leads to misjudgments and unsafe behaviors [18]. Similarly, inappropriate task allocation or excessive task complexity (e.g., during multi-robot supervision) can increase workers’ cognitive load and reduce situational awareness [19]. Therefore, HRC safety should be understood as a systemic issue: performance depends not on optimizing individual components, but on the alignment between human, robot, and task factors. However, existing studies have not yet clarified the mechanisms through which this alignment shapes safety performance, nor have they provided an integrative model to capture these relationships—a gap this study aims to fill.

To theoretically capture this alignment mechanism, the Task–Technology Fit (TTF) theory provides a suitable foundation. The TTF framework conceptualizes performance improvement as a function of the match between task requirements, individual capabilities, and technological functionalities, providing a critical theoretical foundation for this study [20]. It suggests that new technologies enhance performance only when their functionalities fit the needs of users and tasks [21]. Drawing on this logic, this study develops a Human–Task–Robot Alignment (HTRA) framework specific to the construction of HRC, and investigates how it contributes to enhancing safety performance in high-rise HRC, through the following research questions:

• Q1: What are the core components of the HTRA model in high-rise HRC? Specifically, how do the three dyadic interactions (i.e., task–human fit (THA), task–robot alignment (TRA), and human–robot alignment (HRA)) manifest in such scenarios?
• Q2: Does HTRA influence safety performance in HRC? If so, what are the underlying impact mechanisms?

The article is organized as follows. Section 2 introduces the theoretical background of TTF theory and High-altitude HRC safety. Section 3 establishes the research model and related research hypotheses. Section 4 outlines research methods and collects the data sample. Model test results are summarized in Section 5. Section 6 presents a detailed discussion and the implications for theory and practice. Finally, the conclusion offers a summary of findings.

2. Literature Review

2.1. Human–Robot Collaboration (HRC) Safety

Human–Robot Collaboration (HRC) refers to the coordinated interaction between humans and robots in shared workspaces to accomplish specific construction tasks [22,23]. HRC has demonstrated significant potential in reducing labor demand, enhancing productivity, and improving quality control. In high-rise construction scenarios, specialized robots are increasingly used for facade installation, spray painting, cleaning, maintenance and inspection. Examples include: (1) facade installation robots for glass curtain walls [5,6]; (2) curtain-wall-cleaning robots widely deployed (e.g., Nihon Bisoh in Japan used for landmark high-rise structures in Tokyo, Osaka, and Yokohama) [2,3]; and (3) spray-painting robots pioneered by BOZHILIN Co., Ltd. in China [4].

Despite these benefits, high-rise HRC introduces unique safety challenges [9,24]. For instance, small unmanned aerial systems (UAS), while offering mobility and speed, pose significant safety risks if they malfunction or fall from height [12]. Moreover, low-altitude drone operations may distract workers through visual or auditory interference, thereby increasing the likelihood of human errors and subsequent safety incidents [11].

Extensive research efforts have sought to enhance HRC safety, focusing on three main dimensions.

(1)

Robotic optimization for safety.

Most studies concentrate on enhancing robotic situational awareness by optimizing mechanical design and control methods [25,26] to reduce human–robot collision risks [27]. Techniques such as physiological signal decoding [14], gesture and posture recognition [28,29], and skeletal tracking [30,31] help predict human states and intentions, enabling adaptive robot responses. For instance, EEG-based monitoring allowed a bricklaying robot to adjust collaboration speed and distance for safety [14]. Similarly, Sun et al. [10] used physiological data to inform robot motion, improving precision and comfort. Different perception modalities exhibit varying levels of intrusiveness and robustness. Physiological or EEG-based decoding offers rapid responsiveness but may face wearability and latency constraints in dynamic construction environments. In contrast, vision- or pose-tracking systems are more scalable but suffer from degraded performance under occlusion, lighting variation, and visual clutter—conditions typical of high-rise façades. These boundary conditions limit the external validity of robot-side optimizations for high-rise HRC, where wind, height, and line-of-sight interruptions are frequent.

(2)

Task assignment and scheduling

Studies also minimize the shared workspace overlap between humans and robots by task adjustment to mitigate safety risk. Excessive schedule pressure elevates cognitive load and reduces safety [32]. Digital twin–based reinforcement learning supports adaptive task allocation [17], while genetic algorithms optimize ergonomics and robotic capacity, reducing musculoskeletal disorders [33]. Digital twin–reinforcement learning pipelines emphasize adaptive strategies under simulated disturbances and optimize both exposure duration and conflict probability. In contrast, genetic algorithms effectively achieve ergonomic and capacity matching while producing interpretable schedules. Nevertheless, both categories generally assume reliable perception and predictable human responses. In high-altitude construction, where environmental complexity and task uncertainty are high, relying solely on task optimization provides limited safety improvement without concurrent alignment of human and robotic factors.

(3)

Human safety training

Effective training programs aim to equip workers with the knowledge and skills necessary for collaborative tasks [34,35], thereby improving HRC safety. Immersive Virtual Reality (VR) training platforms improve hazard awareness and assess training outcomes through wearable data [15,16]. Such simulations also reveal user experience and risk factors in HRC systems. Although hazard recognition, safety intention, and procedural compliance generally improve in the short term—especially within immersive VR environments—the transfer of these improvements to real sites depends heavily on scenario fidelity, interface transparency, and the degree of fit between task demands and robot behavior. When such alignment is insufficient, training effects tend to diminish, underscoring the need for integrated approaches that jointly consider human, task, and robot adaptation.

Although progress has been made by optimizing the human, robot, and task dimensions individually, misalignment among them can still create risks. In high-rise HRC scenarios, inadequate fits between these elements may generate new safety risks. For instance, technically advanced robots may mismatch workers’ habits or task demands, increasing cognitive load and unsafe behaviors. Likewise, assigning complex tasks to robots with limited perception can cause execution errors and hazardous interactions. These challenges highlight the need for an integrated perspective. Therefore, A fit-oriented perspective that considers interactions among human, robot, and task is essential for improving HRC safety.

In summary, existing studies on HRC safety demonstrate clear progress in three directions: (1) improving robotic perception and control reliability, (2) optimizing task assignment and scheduling to minimize human–robot interference, and (3) enhancing human awareness and capability through training. These approaches represent valuable advances toward safer collaboration. However, they remain fragmented in nature—each focuses on a single dimension under idealized assumptions of stability and predictability. In practice, high-rise HRC involves dynamic environments, multi-agent interaction, and task variability that make isolated optimization insufficient. Consequently, accidents may still arise from misalignment among human, robot, and task factors. This reveals a critical theoretical and practical gap: the need for a systemic perspective that integrates these dimensions and explains how their coordination (or lack thereof) shapes safety performance.

2.2. Factors Influencing Safety Behavior Performance

Behavior-based safety performance is an effective proactive indicator for evaluating overall safety performance [36]. Building on Heinrich’s accident causation theory [37] and subsequent behavioral paradigms, studies primarily conceptualize safety performance through two dimensions: safety compliance (adherence to rules) and safety participation (voluntary safety-enhancing actions) [38].

Research highlights multiple antecedents. At the individual level, factors such as safety awareness and perception, attitudes, workload, and fatigue affect safe actions [39,40]. Organizationally, safety climate, leadership, culture, and management commitment shape motivation to comply or participate [41,42,43].

In HRC, frequent human–robot interaction and the dynamic complexity inherent in collaborative tasks complicate safety behavior performance. Research indicates that frequent interactions with robots substantially influence workers’ behavior, leading to reduced situational awareness, distraction, robot misuse or trust issues [44]. Moreover, dynamic tasks, such as robots replanning paths, changing speeds, or adjusting role assignments, increase cognitive demands.

In non-routine or hazardous situations typical of construction, such as working at heights or operating under time pressure, these behavioral effects become more pronounced. Workers must constantly adapt to unpredictable robot behaviors while maintaining personal safety, which heightens their cognitive load and affects their safety decision-making. A sudden malfunction or trajectory deviation of a robot can trigger instinctive avoidance responses, delay protective actions, or even cause overreliance due to excessive trust. Similarly, insufficient transparency in robot intent (e.g., unclear motion cues or feedback signals) may impair workers’ hazard anticipation, resulting in unsafe proximity or coordination errors. Therefore, the quality of human–robot interaction—trust calibration, communication transparency, and shared situational awareness plays a pivotal role in shaping safety behavior, especially under non-routine and high-risk conditions.

Although recent advances in sensors and computer vision technologies enable real-time detection of risks [45], focusing solely on externally observable unsafe behavior remains insufficient. A comprehensive understanding of the underlying psychological mechanisms, that is, how workers perceive, interpret, and respond to robot behavior, is essential for improving safety. Building on Shin et al.’s [46] worker safety mental process model, this study explores the psychological pathway linking safety perception, safety intention, and behavior performance in HRC, providing a basis for targeted safety interventions.

3. Theoretical Background and Research Hypotheses

3.1. Task–Technology Fit Theory

Task–Technology Fit (TTF) theory was proposed by Goodhue & Thompson [20], originating in the field of information systems. It provides a foundational framework for understanding how well technology aligns with the specific requirements of a task. According to this theory, the effective use and performance benefits of an information system are realized when the system is well-suited to the tasks it is intended to support. A series of research has verified that a good fit between technology, task, and team can lead to better performance outcomes [47,48]. Specifically, performance outcomes depend on the fit between three core constructs: technology characteristics, task requirements, and individual abilities [49]. Liu et al. [50] further conceptualized fit as consisting of three two-way interactions: task–technology fit (TTF), individual–technology fit (ITeF), and task–individual fit (TaIF). In this study, we follow Liu et al.’s framing and model three dyadic alignments (TRA, THA, HRA) to make the interaction mechanisms among task, robot, and human explicit and testable. This approach is particularly suitable because the high interactivity inherent in HRC contexts makes the direct measurement of dyadic interactions more meaningful. Second, distinguishing between different alignments facilitates more precise diagnostics of HRC safety interventions.

Empirical studies have commonly integrated TTF with the Unified Theory of Acceptance and Use of Technology (UTAUT) to explain users’ adoption and usage intentions. In such models, TTF serves as an antecedent that influences performance and effort expectancy, thereby shaping behavioral intention. For instance, research on robot adoption in higher education confirmed that both task–technology and human–technology fit significantly enhance performance expectancy [51]. Similarly, studies on IoT acceptance in the Architecture, Engineering, and Construction (AEC) industry found that TTF strongly predicts practitioners’ willingness to adopt smart construction technologies and interacts with contextual risk factors to explain adoption behavior [52]. Notably, recent studies in the emerging field of construction HRC have significantly expanded the applicability and value of the TTF model. Ma et al. [53] pioneered the study to analyze the “human–task–robot” triangular fit in HRC scenarios and quantitatively assess the substitution potential of construction robots for human workers. However, their research primarily focused on evaluating the development potential of robots and did not examine how the alignment among humans, tasks, and robots affects performance outcomes—particularly safety performance, which is of critical importance in a high-risk HRC context.

While these studies verify the robustness of TTF in explaining technology adoption, they mainly emphasize intention to use in relatively stable environments. In contrast, the construction context is dynamic, uncertain, and safety-critical—where fit not only determines adoption but also directly affects operational reliability and safety outcomes. In HRC, the degree of alignment between humans, tasks, and robots influences workers’ situational awareness, cognitive load, and behavioral safety. Therefore, extending TTF to construction HRC provides a theoretical basis for diagnosing how mismatches among these dimensions lead to unsafe conditions. Building on this common theoretical foundation, this study reconceptualizes TTF into a Human–Task–Robot Alignment (HTRA) framework, focusing on how the triadic alignment among humans, tasks, and robots shapes safety perception, safety intention, and safety behavior in high-rise construction.

3.2. Research Model and Hypotheses

This study proposes a Human–Task–Robot Alignment (HTRA) model and a Bilateral Safety Mental Process (BSMP) model. The former comprises three pairwise alignments (i.e., Task–Robot Alignment, Task–Human Alignment, and Human–Robot Alignment), while the latter encompasses bilateral safety perception, bilateral safety intention, and bilateral safety behavior performance. This integrated theoretical framework is designed to investigate: (1) the impact of HTRA on safety behavior performance; and (2) the underlying mechanisms shaping the mental process of safety behavior performance. Based on these research objectives, the study systematically develops a set of hypotheses, as illustrated in Figure 1.

3.2.1. Task–Robot Alignment (TRA)

Task–Robot Alignment (TRA) refers to the degree of alignment between a robot’s technological capabilities and the demands of the task [54]. Prior research has consistently demonstrated that an incongruence between task demands and technology characteristics can slow decision-making, increase cognitive workload, or lead to a higher likelihood of errors [55,56,57]. Conversely, when a technology possesses the necessary functionalities to effectively perform a given task, it enhances operational efficiency, reduces uncertainty, and improves overall task performance.

In high-rise HRC construction work, the alignment between robot functionalities and task demands is critical for ensuring safety, as these tasks typically involve high physical risks and demand precise execution. Studies have identified key robotic characteristics in high-rise tasks, including automated stability control, real-time hazard detection, and precision handling of construction materials [58,59]. When robots are well-equipped to meet these demands, for example, when curtain wall installation robots possess sufficient accuracy and wind resistance for high-rise HRC [5], workers perceive the robot’s actions and outputs are aligned with the collaborative task, which enhances their confidence in safety and reliability of collaborative construction [60]. Therefore, workers’ confidence in the robot’s ability to execute tasks safely and reliably can effectively reduce their fear of accidents or injuries during collaboration [61,62]. This safety perception further motivates workers to engage in proactive safety behaviors, thereby improving overall safety performance.

Thus, we hypothesize the following:

H1. 

Task–Robot Alignment positively influences safety perception.

3.2.2. Task–Human Alignment (THA)

Task–human alignment (THA) is defined as the congruence between individual competencies and task requirements [63]. It refers to the knowledge, skills, and abilities (KSAs) that employees should possess in order to perform their assigned tasks [64]. HRC introduces new competency requirements for workers, including technical skills for setting robot parameters, proficiency in human–robot interface operations, and adaptability to changing work environments, task planning, and safety management [65].

The fit between employee capabilities and job demands has been widely validated in organizational behavior research as being linked to their work performance and perceived stress level [66]. This is particularly crucial in unstructured high-rise HRC settings, which are typically dynamic, non-routine, and high risk, with task demands shifting in response to constantly changing environments (e.g., human–robot interaction strategies, time allocation) [67]. A strong THA ensures that workers feel capable and in control, thereby enhancing safety perception and promoting safer behaviors. For example, in façade spraying tasks, when sudden wind changes occur, workers who can promptly adjust spraying parameters or assist in nozzle calibration are more likely to maintain smooth collaboration with robots.

Therefore, possessing this necessary technical expertise and the ability to handle unexpected events effectively enhances workers’ safety perception. Conversely, a mismatch between human competencies and task requirements, for instance, insufficient knowledge of the robot’s emergency stop function, can increase psychological stress and compromise safety performance. Based on this, we propose the following:

H2. 

Task–human Alignment positively influences safety perception.

3.2.3. Human–Robot Alignment (HRA)

Human–Robot Alignment refers to the alignment between a worker’s individual capabilities and the technological characteristics of the robot. Better performance outcomes are achieved when individuals possess the necessary knowledge and experience to operate a specific technology [68]. Conversely, poor HRA can increase cognitive load and psychological stress, ultimately diminishing safety perception [35,69,70].

In high-rise HRC, workers are required to possess knowledge (e.g., robotic sensors, control systems, and safety standards), skills (e.g., proficiency with human–robot interfaces, safety management), and abilities (e.g., problem-solving, safety awareness) [65]. Possessing these competencies not only facilitates safe behavior but also helps workers develop a sense of control, the belief that they can effectively manage the robotic system, which enhances their safety perception.

In addition, the role of robot usability in enhancing workers’ safety perception has been widely validated in human–robot interaction studies. Enhanced robot adaptability, real-time error detection, and intuitive feedback and response mechanisms can significantly improve users’ perceived safety and trust in HRC scenarios [71,72]. Similarly, factors such as clear interactive interfaces, motion fluency and predictability, and bidirectional communication are critical for improving safety perception [73]. For example, when cobots are able to interpret human collaborators’ motion intentions and workload, and even adjust their execution speed and proximity accordingly, workers experience reduced mental workload and an enhanced sense of safety [74].

Therefore, the following hypothesis is proposed:

H3. 

Human–Robot Alignment positively influences safety perception.

3.2.4. Safety Constructs

Research on the antecedents of safety behavior has extensively validated and established the psychological process from perception to intention to behavior [46,75]. However, existing construction safety studies predominantly focus on human safety, often overlooking the safety risks associated with robots, despite their evolving role from traditional mechanical equipment to collaborative counterparts in construction settings. Robots, characterized by heightened autonomy, intelligence, high cost and high precision, play a critical role in the HRC team. Therefore, this study introduces the concept of bilateral safety, expanding the traditional safety framework from unilateral protection (solely addressing human safety) to bilateral protection (ensuring the safety of both human workers and robotic counterparts). This perspective is operationalized through three key constructs: bilateral safety perception, bilateral safety intention, and bilateral safety behavior.

Bilateral safety perception (BSP) is defined in this study as the extent to which workers feel relaxed, safe and comfortable in HRC tasks [76], encompassing both self-protection and concern for the robot’s safety. Safety perception can significantly shape behavioral intentions according to the Theory of Planned Behavior (TPB) [75,77]. For example, during high-rise curtain wall installation, if workers perceive task complexity and robot reliability as controllable, they not only consider themselves to be in a safe state but also believe that their robotic counterparts will not be damaged by vibrations or shaking. This positive bilateral safety perception then translates into a clear bilateral safety intention, whereby workers consciously protect both themselves and the robots from harm.

H4. 

Bilateral safety perception positively influences bilateral safety intention.

Bilateral safety intention is defined as a worker’s deliberate decision to engage in safety behaviors that protect both themselves and their robotic teammates; it reflects the worker’s subjective safety willingness and determination. According to the Theory of Reasoned Action (TRA) [78], intention is the most immediate predictor of behavior. Numerous studies have confirmed the strong link between safety intention and actual safety behavior, emphasizing that workers who intend to follow safety protocols are more likely to do so in practice [79]. In high-rise HRC, workers with strong bilateral safety intentions tend to demonstrate bilateral safety behaviors during collaboration, that is, valuing their own safety while simultaneously taking actions to protect the robot from potential damage.

H5. 

Bilateral safety intention positively influences bilateral safety behavior.

4. Research Method

4.1. Questionnaire Design

Given the nascent stage of construction robot applications, rigorous data collection necessitates both precise questionnaire design and stringent selection of respondents with relevant experience. The measurement items were primarily established and adapted based on the mature investigation paradigms of TTF and safety behavior performance. To contextualize existing constructs within high-altitude HRC scenarios, this study first implemented observational tracking at construction projects utilizing high-altitude robotics, thereby identifying critical characteristics and requirements across tasks, workers and robots. Additionally, for bilateral safety constructs (i.e., BSP, BSI, BSB), questionnaire items were augmented to incorporate robotic safety considerations. For instance, traditional safety compliance items typically include statements like the following:

• I always use appropriate personal protective equipment (e.g., safety harness, helmet, gloves).
• I strictly follow safety procedures to ensure my own and other workers’ safety.

In this study, the items were expanded to include robot safety, such as the following:

• I always equip the robot with appropriate protective measures (e.g., collision protection, cushioning or buffer pads).
• I strictly follow safety construction procedures to ensure the robot is not damaged.

Following the appropriate modifications, we conducted a series of focus group discussions with five experts experienced in construction robotics and automation (two Ph.D. candidates, two post-doctoral researchers, and one employee from a construction enterprise). Each participant had 3–5 years of experience in robot-related projects. Based on their feedback, three revisions were made to improve the questionnaire’s clarity and applicability: (1) the wording of several overly academic or ambiguous items was simplified to enhance readability and practical relevance; (2) the subject of questions was changed from “onsite workers” to “you (or workers)”; (3) an open-ended question was added to capture practical experiences, “What kinds of human–robot collaboration safety events have you witnessed or experienced on site, and what were the main causes?

Furthermore, during these discussions, the experts emphasized that the presence or absence of standardized construction procedures and robot-specific standards could significantly affect the effectiveness of HRC. Their insights led to the inclusion of two additional diagnostic items in the questionnaire: (1) “In the projects you have participated in (or are familiar with), are there any fixed procedures for construction robots (e.g., operation sequences, construction processes, or workflows)?” (2) “In the projects you have participated in (or are familiar with), are there any construction standards specifically formulated for construction robots?” These items were later used to test the existence of HRC-specific procedures as potential control variables in the empirical model.

Measurement items were adapted from several mature scales to ensure both theoretical validity and contextual applicability. Specifically, items for task–technology alignment were derived from the Task–Technology Fit (TTF) framework [80,81,82]; items for safety behavior performance were adapted from the construction safety behavior scales developed by Neal and Griffin [79]; and items related to safety perception and intention were informed by Liu et al. [83] and subsequent HRC safety studies. Minor linguistic and contextual modifications were made to reflect high-altitude HRC operations.

The anonymous questionnaire commences with a clear declaration of research objectives and comprises two sections—participant background information and construct evaluation. It contains 3 demographic items and 29 measurement items assessing 6 theoretical constructs and 2 control variables. As presented in

Table 1. Measurement items for constructs.

Construct	Item	Codes	Measurement	Sources
HRA	Human–Robot Alignment	HRA1	I collaborate smoothly with the robot during construction tasks.	[80]
		HRA2	I work efficiently with robots on site.
		HRA3	I can rely on the robot to complete assigned tasks.
		HRA4	I trust the information and feedback provided by the robot.
TRA	Task–Robot Alignment	TRA1	The facade spraying robot meets the safety requirements for high-altitude operations.	[81,82]
		TRA2	The facade spraying robot improves the efficiency of high-altitude operations.
		TRA3	The facade spraying robot meets the precision requirements for high-altitude tasks.
THA	Task–Human Alignment	THA1	I possess the necessary skills for facade spraying tasks (e.g., robot operation or assistance).	[84]
		THA2	I possess the necessary abilities for facade spraying tasks (e.g., identifying on-site risks).
		THA3	I have relevant experience in facade spraying tasks (e.g., handling emergencies).
		THA4	I have a strong safety awareness and sense of responsibility in facade spraying operations.
BSP	Bilateral Safety Perception	BSP1	Most of the time, I feel safe when working with robots on site.	[76]
		BSP2	Most of the time, I believe the robot will not be damaged during the collaboration.
		BSP3	Most of the time, I feel relaxed and not nervous or fearful when working with robots.
		BSP4	Most of the time, I do not worry about the robot being damaged during operations.
		BSP5	I always wear appropriate personal protective equipment to protect myself during human–robot collaboration (e.g., safety harness, helmet, gloves).
		BSP6	I always ensure that the robot is equipped with suitable protective devices to prevent damage or collisions (e.g., bumpers, cushioning pads).
BSI	Bilateral Safety Intention	BSI1	I intend to always follow safety procedures during high-altitude work to protect myself and coworkers.	[83]
		BSI2	I intend to always follow safety procedures to avoid damage to the robot during high-altitude tasks.
		BSI3	I intend to ensure maximum safety for myself during future high-altitude operations.
		BSI4	I intend to prevent robot damage to the greatest extent during future high-altitude operations.
		BSI5	I intend to actively participate in safety training or meetings to enhance worker safety.
		BSI6	I intend to actively participate in training or meetings to prevent robot-related damage.
BSB	Bilateral Safety Behavior Performance	BSB1	I strictly follow safety procedures to prevent harm to myself and other workers.	[79]
		BSB2	I strictly follow safety procedures to prevent damage to the robot.
		BSB3	I actively participate in safety training to learn how to protect myself and my coworkers.
		BSB4	I actively participate in safety training to learn how to prevent damage to the robot.
		BSB5	I proactively suggest ways to improve worker safety on site.
		BSB6	I proactively suggest ways to improve robot safety during operations.

, all finalized measurement items employed a five-point Likert scale (1 = Strongly Disagree to 5 = Strongly Agree) to capture response intensity. The background section incorporated validation checks to ensure respondents possessed the requisite experience with robotic systems in high-altitude environments.

4.2. Data Collection and Analysis

The formal questionnaire was distributed through an online questionnaire platform, Wenjuanxing (www.wjx.cn, accessed on 15 February 2025), employing the snowball sampling method [85,86]. Four categories of respondents were prioritized for this study: (1) construction professionals involved in projects utilizing high-altitude robots or other construction robots, (2) technical personnel from construction robotics R&D teams, (3) construction robots operation trainers, and (4) academics and researchers specializing in construction automation. Initially, we contacted managers of construction enterprises and researchers who had implemented robotic systems and invited them to participate in the survey. These managers then forwarded the questionnaire link to other practitioners engaged in human–robot collaboration within their organizations or partner projects. This recruitment method was chosen because enterprises currently applying robots in high-rise construction remain limited, and the target population is highly specialized and dispersed. This approach ensured that all respondents possessed direct experience with HRC in construction.

A detailed definition of high-altitude operation robots and descriptions of their application scenarios were provided to adequately supplement the understanding of the technological features and application contexts. To ensure data quality, invalid responses were excluded if (1) the completion time was less than 90 s or greater than 600 s, (2) any items contained missing values, or (3) identical answers were provided across all items. Between December 2024 and March 2025, a total of 84 valid responses were collected.

Partial Least Squares Structural Equation Modeling (PLS-SEM) was utilized in this study to validate the research model and corresponding hypotheses via the SmartPLS software package (version 4.1.0.4). This method is well-suited for exploratory studies with bootstrapping samples. Since this study aimed to investigate how human–task–robot alignment and interactions impact safety performance, it was founded on an exploratory nature. Moreover, PLS-SEM shows high statistical power when handling a small sample size. These two advantages enable PLS-SEM to be prevalent in many construction management studies [87], particularly for safety performance [43,88]. PLS-SEM was adopted first to assess the measurement model with reliability and validity tests and then to estimate the structural model with significance tests of the theoretical paths.

5. Results

5.1. Descriptive Results

Table 2. Details of respondents.

Characteristic	Category	Frequency	Percentage
Organization	Construction company	40	47.62%
	Project owner/client	8	9.52%
	Robot supplier	22	26.19%
	University or research institution	14	16.67%
Work experience	1–5 years	44	52.38%
	6–10 years	28	33.33%
	11–15 years	9	10.72%
	>15 years	3	3.57%
Role	Project manager	22	26.19%
	Robot operator	13	15.48%
	Traditional construction worker (HRC assistant)	5	5.95%
	On-site safety manager	8	9.52%
	Robotics development engineer	27	32.14%
	Researcher	9	10.72%

presents the demographic information of the survey respondents. The participants came from diverse organizational backgrounds, including construction companies (47.6%), project owners/clients (9.5%), robot suppliers (26.1%), and universities or research institutions (16.6%). Many respondents reported less than 10 years of work experience, with 52.3% having 1–5 years and 33.3% having 6–10 years. This distribution is consistent with the relatively recent introduction of construction robotics into practice, underscoring that HRC remains an emerging field. In terms of professional roles, respondents included project managers (26.2%), robot operators (15.5%), traditional construction workers assisting HRC tasks (5.9%), on-site safety managers (9.5%), robotics development engineers (32.1%), and researchers (10.7%). This distribution indicates that the sample covers both managerial and technical personnel, as well as practical frontline workers, ensuring a balanced perspective on human–robot collaboration safety.

5.2. Measurement Model Test

PLS-SEM (Partial Least Squares Structural Equation Modeling) has established a set of empirical guidelines for evaluating both reflective and formative measurement models [89]. The theoretical model in this study was developed as a reflective measurement model. Accordingly, four core evaluation criteria were applied to the reflective indicators: indicator reliability, internal consistency reliability, convergent validity, and discriminant validity. These assessments are essential to ensure the robustness of the measurement model.

Specifically, any inconsistent or insignificant indicators were considered for removal based on the following criteria: (1) constructs and their indicators with Cronbach’s alpha or composite reliability below 0.7 were considered to lack internal consistency [90]; (2) indicators with factor loadings below 0.4 were deemed inconsistent, indicating poor convergent validity [91]; (3) constructs with AVE values less than 0.5 were considered unacceptable, suggesting insufficient convergent validity [92] and (4) the outer loading of an indicator on its assigned construct should exceed its cross-loadings on other constructs; additionally, the square root of each construct’s AVE should be greater than its correlations with other constructs in the model to establish sufficient discriminant validity [90].

The results indicated all indicators satisfied the required evaluation standards. The relevant evaluation results are detailed in

Table A1. Evaluation criteria of the measurement model.

Construct	Indicator	Indicator Reliability			Internal Reliability	Consistency	Convergent Validity
		Loading	t-Value	p-Value	Cronbach’s a	CR	AVE
BSB	BSB1	0.885	21.369	0.000	0.935	0.946	0.690
	BSB2	0.871	15.054	0.000
	BSB3	0.852	17.713	0.000
	BSB4	0.861	22.330	0.000
	BSB5	0.866	27.429	0.000
	BSB6	0.878	22.271	0.000
BSI	BSI1	0.838	14.996	0.000	0.933	0.947	0.749
	BSI2	0.934	33.636	0.000
	BSI3	0.867	16.755	0.000
	BSI4	0.782	9.861	0.000
	BSI5	0.875	16.308	0.000
	BSI6	0.889	18.055	0.000
BSP	BSP1	0.784	8.702	0.000	0.811	0.877	0.643
	BSP2	0.879	22.211	0.000
	BSP3	0.722	4.109	0.000
	BSP4	0.848	19.762	0.000
HRA	HRA1	0.823	20.022	0.000	0.891	0.925	0.755
	HRA2	0.848	34.112	0.000
	HRA3	0.853	12.063	0.000
	HRA4	0.804	9.332	0.000
THA	THA1	0.823	15.655	0.000	0.852	0.900	0.693
	THA2	0.848	25.632	0.000
	THA3	0.853	30.346	0.000
	THA4	0.804	4.977	0.000
TRA	TRA1	0.885	6.148	0.000	0.848	0.908	0.767
	TRA2	0.840	7.471	0.000
	TRA3	0.901	5.778	0.000

and

Table A2. Fornell-larcker criterion (latent variable correlations) for discriminant validity evaluation.

	BSB	BSI	BSP	HRA	THA	TRA
BSB	0.830 *
BSI	0.739	0.865 *
BSP	0.705	0.635	0.802 *
HRA	0.706	0.618	0.664	0.869 *
THA	0.674	0.703	0.745	0.778	0.879 *
TRA	0.682	0.593	0.741	0.763	0.738	0.876 *

Note: * The square root of average variance extracted (AVE).

. Thus, the final measurement items were appropriate for their respective constructs and demonstrated satisfactory reliability and validity for structural model evaluation. Thus, the measurement items were appropriate for their respective constructs and reliable and valid for the structural model evaluation.

5.3. Structural Model Test

5.3.1. Original Structural Model

Figure 2 and

Table 3. Criteria results of the original structural model.

Hypothesis	Path	Coefficient	t-Value	p-Value	f2	R2	Result
H1	HRA > BSP	0.313	3.536	p < 0.0001	0.429	0.944	Supported
H2	TRA > BSP	0.274	2.823	p < 0.0005	0.169		Supported
H3	THA > BSP	0.424	3.294	p < 0.0001	0.434		Supported
H4	BSP > BSI	0.635	6.708	p < 0.0001	0.675	0.453	Supported
H5	BSI > BSB	0.739	9.586	p < 0.0001	1.203	0.546	Supported

show the results of the hypothesis testing. The structural model’s quality was assessed using R², indicating the model’s predictive power, and f², measuring each predictor’s contribution to variance explanation [93]. Human–task–robot alignment plays a significant role in improving the bilateral safety perception, intention, and performance. The results demonstrate that the three HTRA components (i.e., TRA, THA and HRA) exhibit reasonable predictive accuracy for improving safety perception, with H1 (HRA → BSP) and H3 (THA → BSP) showing strong effects, and H2 (TRA → BSP) showing a relatively moderate effect (0.274, f² = 0.169). External impact also exhibited significantly positive effects, i.e., H4 (0.635, p < 0.0001) and H5 (0.739, p < 0.0001). The coefficient of determination (R²) reflects the proportion of variance explained by exogenous constructs. Following Hair et.al [90], R² values of 0.75 and 0.50 are typically interpreted as substantial and moderate. Furthermore, all VIF values were below 5, confirming the absence of multicollinearity. The model also exhibited good overall fit, with SRMR = 0.080 (≤0.08) and NFI = 0.911 (>0.90), indicating that the structural model provides an acceptable and reliable representation of the observed data.

5.3.2. Control Variable Experiment

This study incorporates the critical attribute of regulation and standardization by introducing control variables and conducting a comparative experiment to evaluate the impact of standardized provisions on human–task–robot coordination (see

Table A3. Results of control variables experiment.

Path	Coefficient	t-Value	p-Value
HRA > BSP	0.313	3.528	0.000
TRA > BSP	0.274	2.823	0.004
THA > BSP	0.424	3.294	0.000
BSP > BSI	0.635	6.709	0.000
BSI > BSB	0.739	9.586	0.000
Standards > HRA	0.646	2.277	0.022
Standards > RTF	0.250	0.758	0.448
Standards > THA	0.181	0.549	0.583
Process > HRA	0.650	n/a	n/a
Process > TRA	0.404	n/a	n/a
Process > THA	0.468	n/a	n/a

Note: n/a = not applicable.

for complete results). Two variables were included: HRC-specific standards and HRC-specific construction procedures. The former refers to formal regulatory documents and technical specifications formulated specifically for HRC (e.g., safety operation procedures, robot usage requirements, task technical specifications, and risk prevention protocols). The latter is defined as specific construction processes and methods tailored to the characteristics of HRC operations and robotic technologies (e.g., process design, task allocation, and detailed human–robot interaction requirements).

Theoretically, HRC-specific construction processes are closely related to safety performance. However, due to the early stage of HRC development, standardized construction processes specific to HRC have not yet been widely formulated or adopted in practice, and the majority of respondents reported their absence (i.e., the variable remained constant across all observations). Therefore, this variable contributed no explanatory variance, and SmartPLS did not report a t-value or p-value for its coefficient. As shown in Figure 3, the final experimental results indicate a statistically significant association between HRC standards and Human–Robot Alignment (HRA), whereas no significant effects were observed on Task–Human Alignment (THA) or Task–Robot Alignment (TRA). No significant correlations were found between HRC-specific construction procedures and the three HTRA indicators.

6. Discussion

6.1. HTRA Model

The positive effect of TRA (H1) supports prior findings that the robot’s ability to match task requirements is particularly critical in dynamic, uncertain, and high-risk construction environments. These capabilities—such as sufficient reliability and autonomy, advanced environmental perception, and flexible execution [94,95], and even the ability to perform imitation learning for specific tasks and respond to unexpected scenarios [96]—can significantly influence operator trust, making human workers perceive the interaction with robots as safe [97]. Furthermore, from a managerial perspective, Liang et al. [98] suggest that robot deployment should be optimized according to task characteristics and workspace configuration, which helps minimize unnecessary interactions between humans and robots.

Similarly, THA (H2) reflects the critical role of human competencies in adapting to complex, non-routine high-altitude HRC tasks. When workers possess the knowledge, skills, and experience required to perform specific tasks (i.e., working at height and collaborating with robots), they demonstrate greater confidence in managing tasks and mitigating risks, thereby enhancing their safety perception. This finding supports prior HRC research that emphasizes the importance of immersive virtual environment-based training in improving workers’ safety perception for collaborative tasks [10,99].

HRA (H3) is the most influential factor among the HTRA factors on safety perception, confirming prior research in ergonomics and human–robot interaction research. The impact of this alignment on safety perception is evident in two key dimensions. First, at the interaction interface level, improved mutual understanding between human and robot (e.g., behaviors, states, and intentions) improves situational awareness and trust, thereby enabling safer and closer collaboration. Second, robotic attributes (e.g., motion speed, proximity, and behavioral predictability) significantly enhance workers’ psychological safety by improving physical comfort and perceived controllability during collaborative tasks [10,74].

Notably, the three alignments exhibit varying degrees of contribution to bilateral safety perception, with Task–Robot Alignment (TRA) proving less influential compared to HRA and THA. Although the path coefficient of TRA → BSP (β = 0.274, p < 0.005) is smaller, it remains statistically significant and practically meaningful. This result reflects the current stage of HRC development, where task–robot coordination is often limited by technological immaturity, restricted autonomy, and non-standardized interaction interfaces. Thus, while TRA contributes positively to workers’ bilateral safety perception, its effect is moderate compared with human-centered alignment dimensions. As robotic systems become more transparent, adaptive, and better integrated with task requirements, the practical influence of TRA on safety perception is expected to increase accordingly.

Moreover, this variation is particularly evident in high-rise construction environments, where spatial constraints, visual occlusion, and environmental disturbances (e.g., wind, glare, vibration) intensify operational uncertainty. Under such conditions, the performance of robots is highly sensitive to task complexity and environmental noise, making human adaptability, perception, and decision-making indispensable to maintaining safety. Workers’ ability to anticipate contextual risks, adjust cooperation timing, and compensate for robotic limitations plays a decisive role in preventing accidents.

This finding of HRC fields also reflects the early developmental stage of construction robotics, where most robotic systems still exhibit limited autonomy, weak environmental perception, and non-standardized interfaces. Consequently, human factors—such as situational awareness, experience, and trust calibration—play a dominant role in ensuring safe coordination. This explains why, in the current stage, human-centered alignments (HRA and THA) contribute more strongly to safety perception than Task–Robot Alignment (TRA). This challenges prior studies [50] which suggest that Task–Technology Alignment has a superior effect on performance than the other alignments. The apparent discrepancy can be attributed to contextual and technological differences between the two studies. Liu et al. [50] investigated highly standardized manufacturing systems, where robots operate in controlled environments with stable task sequences and established safety interfaces—conditions that enable technology-dominant alignment. The comparative analysis between these contexts reinforces that the relative influence of alignment dimensions is context-dependent and evolves with technological maturity. As construction robotics progresses toward higher autonomy and better integration, the influence of TRA on safety performance is expected to increase correspondingly.

6.2. Bilateral Safety Behavior Performance Model

The empirical validation of H4 and H5 confirms the effectiveness of the proposed bilateral safety behavior mental process model in the context of construction HRC. The results demonstrate a significant sequential path from bilateral safety perception to bilateral safety intention, and ultimately to safety behavior performance. This dynamic psychological mechanism reveals the cognitive-behavioral processes that underlie workers’ actions in human—robot teaming environments.

The positive influence of bilateral safety perception on safety intention highlights a cognitive–motivational mechanism central to collaborative safety behavior. When workers perceive that both they and the robot are operating safely through transparent communication, predictable actions, and stable task execution, they experience reduced uncertainty and a stronger sense of control. This mutual safety perception builds trust and shared responsibility, motivating workers to intentionally maintain safe interactions and to prevent risky events before they occur. In essence, perception transforms into intention as workers internalize safety as a joint goal rather than an individual obligation.

In construction HRC, this process has important practical implications. High-altitude operations often involve dynamic hazards, fluctuating environmental conditions, and close human–robot proximity. Under these circumstances, stronger bilateral safety intentions encourage workers to maintain appropriate distances, monitor robot motions proactively, and coordinate real-time responses, thereby creating a self-reinforcing safety feedback loop. This mechanism helps transform passive compliance into active collaboration, enhancing both human and robot safety in complex, high-risk construction scenarios.

Furthermore, the concept of bilateral safety proposed in this study extends the classical Heinrich’s Accident Causation Theory, which traditionally emphasized human unsafe acts and conditions as primary precursors to accidents. In HRC contexts, the accident chain becomes bidirectional, encompassing both human and robotic agents as interactive components of a socio-technical system. Human errors may compromise robot safety (e.g., task miscoordination or miscommand), while robot malfunctions or unpredictable actions may endanger human operators.

By introducing bilateral safety perception, intention, and behavior, this study expands Heinrich’s framework from a human-centered causality model to a human–robot co-causality paradigm, emphasizing mutual safety responsibility and interactive risk activation within collaborative environments. This theoretical expansion reflects the shift from unidirectional accident causation—where safety outcomes are driven primarily by human actions—to co-dependent risk chains, in which human and robotic agents dynamically influence each other’s safety status.

Under this framework, workers first assess the safety status of their collaboration with robots by gathering multiple types of information and subsequently form safety-related appraisals that guide their decision-making. Once workers form a concrete intention (e.g., maintaining a safe distance or actively monitoring robot behavior), they are more likely to engage in proactive safety actions [46]. The bilateral nature of the construct thus enriches the traditional safety behavior model by emphasizing that construction workers consider both human and robot safety outcomes before forming an intention and initiating safety actions.

These results further illustrate an evolving cognitive perspective in HRC, where the robot’s operational status and behavior significantly influence workers’ safety judgments and behavioral decisions. In such settings, workers continuously assess not only the risks to their own well-being but also the operational safety of their robotic teammates. This perception of the responsibility of bilateral safety represents a cognitive shift from viewing robots merely as tools to recognizing them as intelligent, interdependent partners in achieving safe and efficient collaboration.

6.3. Implication for Theory and Practice

This study offers several theoretical contributions to the understanding of Task–Technology Fit (TTF) in the context of human–robot collaboration (HRC) safety. First, grounded in TTF theory, this research develops a novel theoretical framework for evaluating human–task–robot alignment specific to construction HRC scenarios, thereby broadening the conceptual scope of HTRA within high-risk, technology-integrated environments. Additionally, by extending the concept of bilateral safety, the study introduces and operationalizes a set of constructs—bilateral safety perception, bilateral safety intention, and bilateral safety behavior performance—which enhances the applicability of traditional safety models in emerging HRC-enabled construction settings.

Second, the findings empirically confirm that higher levels of alignment among humans, tasks, and robots are associated with improved safety outcomes. This insight addresses a key research gap, as prior studies on HRC safety have predominantly emphasized technological advancements, often overlooking the importance of alignment between collaborative agents.

Lastly, by incorporating control variables and conducting comparative experiments, the study evaluates the influence of standardization as a critical project attribute for effective human–robot coordination. Specifically, standardized construction procedures and formalized HRC operational guidelines were found to significantly improve human–robot alignment, facilitating smoother interaction and safer working conditions.

The findings of this study also provide a new perspective for management practices, shifting from isolated improvements to systemic optimization. First, HRC managers should recognize the critical importance of alignment among humans, robots, and tasks, rather than treating each as an isolated or unidimensional construct. Construction firms can utilize the proposed matching framework to assess and optimize collaborative workflows and develop targeted organizational strategies tailored to HRC operations.

Second, the survey results show that the lack of standardized HRC construction workflows remains a pressing challenge, particularly in terms of aligned task allocation and collaborative workflow design. In existing HRC systems, the division of roles and responsibilities between construction workers and robots remains ambiguous. As a result, workers often continue to follow traditional construction practices, which are incompatible with robot-integrated operations. This misalignment leads to operational confusion and introduces additional safety risks on-site. Therefore, collaborative work standards should be developed as integrated frameworks comprising the following core modules:

(1)

Human–robot role division and authority boundaries, specifying which tasks are executed autonomously by robots and which require human supervision or joint control.

(2)

Interface and communication protocols, defining how information, warnings, and task updates are exchanged between human operators and robotic systems.

(3)

Emergency response and recovery procedures, ensuring rapid and coordinated actions in case of system failure, near misses, or hazardous events.

Third, to enhance Human–Robot Alignment (HRA) in practice, construction enterprises should invest in targeted training programs and technological transformation initiatives. Training should emphasize situational awareness, shared control, and real-time communication between human operators and robots, while technical upgrades should focus on improving robot transparency, motion predictability, and adaptive task allocation mechanisms. Together, these strategies can enhance mutual understanding and ensure both human and robot safety during complex construction operations.

In summary, these implications underline the importance of developing standardized, modular, and adaptive HRC management systems that integrate human, technological, and procedural dimensions. Such systems can promote the safe, efficient, and scalable adoption of HRC in high-rise and other complex construction scenarios.

7. Conclusions

This study proposed a Human–Task–Robot Alignment (HTRA) model and a Bilateral Safety Behavior model to examine how alignment mechanisms shape safety performance in the construction of HRC. Using 84 valid survey responses analyzed via Partial Least Squares Structural Equation Modeling (PLS-SEM), the results confirm that the alignment between humans, robots, and tasks (i.e., HRA, THA, and TRA) serves as a pivotal mechanism enhancing bilateral safety perception, which subsequently promotes safety intention and behavior.

Importantly, at the current stage of HRC development, human-centered alignments (HRA and THA) have a stronger effect on safety perception than Task–Robot Alignment, reflecting the dominant role of human adaptability and decision-making in complex, non-standardized construction environments. Moreover, standardized HRC guidelines were found to significantly improve Human–Robot Alignment, underscoring the necessity of clear workflows, interface protocols, and task coordination standards.

While this study focuses on high-rise construction, the underlying mechanisms identified, including the importance of human adaptability, clear role division, and standardized collaboration procedures, are equally relevant to other construction contexts, such as prefabrication, infrastructure inspection, or on-site assembly. In these environments, the HTRA framework can guide managers in optimizing team–robot coordination, improving situational awareness, and developing adaptive safety management strategies suited to varying levels of automation and task complexity.

There are also several limitations. The relatively small sample size (n = 84) reflects the current scarcity of construction projects employing robots—particularly for high-altitude operations. Nevertheless, the sample meets the minimum requirement for PLS-SEM analysis and provides valid insights into an emerging field. As the adoption of construction robots increases and empirical data accumulates, future studies should validate and refine the model using larger, more diverse samples and cross-cultural comparisons to enhance its generalizability. Second, this study focused on high-rise construction at an early stage of HRC development, where robotic autonomy and standardization are limited. As technology matures, comparative studies across various automation levels and cross-cultural contexts become valuable to test whether the influence of alignment dimensions shifts with technological and organizational evolution.

Future research is encouraged to further validate and expand this model across a wider spectrum of HRC scenarios, involving diverse task types (e.g., assembly, inspection, and maintenance), varying levels of automation, and heterogeneous workforce compositions. It will be valuable to examine whether these results remain consistent across varied contexts. To address current limitations, future work should incorporate multimodal data (e.g., physiological signals, motion trajectories, and eye-tracking data) to complement self-reported measures and capture real-time safety perceptions and behaviors. Field experiments or longitudinal case studies within ongoing HRC projects can provide deeper insight into how alignment and safety cognition evolve over time as robots and workflows mature. Moreover, this study serves as an exploratory effort situated in the pre-paradigm stage of HRC development. As HRC tasks and workflows become better-defined and robotic technologies advance, Task–Robot Alignment may become increasingly attainable as human involvement decreases. Under these evolving conditions, further validation is necessary to assess whether influence patterns remain stable over time or shift as collaboration paradigms evolve.