Every Second Counts: A Comprehensive Review of Route Optimization and Priority Control for Urban Emergency Vehicles

Abstract

Emergency vehicles (EMVs) play an important role in saving human lives and mitigating property losses in urban traffic systems. Due to traffic congestion and improper priority control strategies along the rescue route, EMVs may not be able to arrive at rescue spots on time, which also increases traffic risk and has a negative impact on social vehicles (SVs). The greater the negative impact on SVs, such as increased delay times and queue length, the more profound the negative impacts on urban environmental sustainability. Proper rescue route selection and priority control strategies are essential for addressing this problem. Consequently, this paper systematically reviews the studies on EMV routing and priority control. First, a general bibliometric analysis is conducted using VOSviewer. This study also classifies the existing studies into three parts: EMV travel time prediction (EMV-TTP), EMV routing optimization (EMV-RO), and EMV traffic priority control (EMV-TPC). Finally, this study provides future research suggestions on five aspects: 1. uncovering authentic demand characteristics through EMV data mining, 2. incorporating the distinct characteristics of EMV in EMV-RO models, 3. implementing active EMV-TPC strategies, 4. concentrating more on the negative impacts on SVs, and 5. embracing the emerging technologies in the future urban traffic environment.

1. Introduction

The response time of EMVs (e.g., police cars, fire trucks, ambulances, engineering rescue vehicles, military vehicles) is a critical evaluation indicator of a city’s emergency management capabilities and sustainability [1,2,3]. For most EMVs, the term “response time” denotes the time interval between an EMV receiving the rescue alarm and reaching the rescue spot, as shown in Figure 1. The emergency medical service (EMS) is an exceptional case as it also includes the on-scene time and return time, which depend on the patient’s condition.

Unlike SVs, measly delays in EMV response times can have tremendous consequences. For instance, it is estimated that a one-minute reduction can increase by up to 24% the survival chance of out-of-hospital cardiac arrests [4] and save 1542 USDs in the cost of treatment per cardiac arrest patient [5]. Furthermore, research indicated that a 10% extension in the response time could lead to a 4.7% reduction in the likelihood of resolving a crime [6].

According to the findings of existing studies, the factors affecting the EMV response time include travel distance, departure time, number of busy EMVs, traffic congestion, and heavy rainfall [7,8]. Among these factors, traffic congestion has been identified as a critical factor impacting EMV response times and rescue reliability, especially during the morning and evening peak hours [7,9]. EMVs during rescue tasks are not restricted to driving routes, directions, speeds, and traffic signals, which results in higher traffic accident risk than that for similarly sized SVs [10]. Consequently, numerous studies have proposed many algorithms and strategies to improve EMV rescue efficiency [11], safety [12], and reliability [13] as much as possible, including EMV-RO algorithms and EMV-TPC strategies, which are the most adopted methodologies in existing studies. EMV-RO algorithms can optimize EMV rescue routes to reduce response times and improve reliability by avoiding traffic jams. EMV-TPC strategies can maintain EMV expected speeds and reduce driving risks by adjusting traffic signal controls and road link traffic management.

In the literature, a few papers have reviewed studies on reducing EMV response times from different perspectives [14,15,16,17,18]. Umam et al. reviewed strategies for decreasing response times in terms of locating, allocating, dispatching, and others [14]. Humagain et al. concluded methods for reducing response times and categorized them into route optimization and priority control techniques [15]. Kamble and Kounte classified the traffic signal priority control models into three categories: routing-based, scheduling-based, and miscellaneous [16]. Yu et al. sorted traffic control strategies into three groups: optimization of routes, signal priority control, and lane reservation [17]. Lakshmi et al. [18] categorized the ambulance rescue route optimization problem into two types: from the ambulance station to the injured location and from the injured location to the nearby medical center. While these review papers are informative and provide valuable reviews, they still have limitations, summarized as follows:

• These review papers ignored significant differences between various EMVs and SVs regarding optimization objectives, vehicle attributes, vehicle dynamics, and driving policies, which may deviate from reality [14,17,18].
• Due to ignorance of the differences between EMVs and SVs, insufficient emphasis was placed on EMV-TTP methods, which are the foundation for achieving optimal EMV-RO outcomes [14,15,16,17,18].
• These review papers mainly emphasized EMV-TPC at intersections and focused less on road links (e.g., giving way, setting up EMV-exclusive lanes, utilization of opposite lanes and shoulder lanes), which cannot ensure the expected speed along the whole route [14,15,16,18].
• These reviews had limited consideration of the application of emerging technologies such as unmanned aerial vehicles (UAVs) and Vehicular Ad-hoc Networks (VANETs) in the implementation of emergency vehicle route planning and along-route traffic priority control strategies [14,15,16,17,18].

With the rapid development of emerging technologies (e.g., Big Data, UAVs, Artificial Intelligence, and VANET) [19], traditional EMV rescue processes in urban traffic systems are undergoing disruptive transformations. These technologies not only significantly improve EMV performance but also pose higher theoretical and practical requirements on the EMV-TTP methods, EMV-RO algorithms, and EMV-TPC strategies.

Therefore, oriented by EMV authentic rescue demand characteristics, this study provides a comprehensive review of the related literature regarding EMV routing optimization and en-route traffic priority control strategies, including EMV-TTP methods, EMV-RO algorithms, and EMV-TPC strategies, which are three hotspots that have captured the most attention in existing studies. The EMV-TPC strategies include EMV-TPC at intersections and road links. Furthermore, this study focuses on the theoretical and practical aspects of the EMV routing optimization algorithms and various priority control strategies under the emerging intelligent and connected environment, as depicted in Figure 2.

This study’s results make the following contributions:

• We employed a comprehensive search methodology to systematically retrieve literature on EMV-TTP methods, EMV-RO algorithms, and EMV-TPC strategies up until March 2024. We used VOSviewer software 1.6.18 to visualize keyword co-occurrences and the most cited journals in the collected studies to reveal the current research trends and hot topics.
• In the context of emerging technologies such as VANETs and UAVs, we specifically reviewed related research on EMV-TTP, including model-based, simulation-based, API-based, and UAV-based methods. As the crucial input parameters for EMV-RO and EMV-TTP, enhancing EMV-TTP outcome accuracy can significantly improve the precision and feasibility of EMV-RO and EMV-TTP.
• Considering the significant differences in the characteristics between various EMVs and SVs regarding optimization objectives, vehicle attributes, vehicle dynamics, and driving policies, we divide the EMV-RO into three subclassifications for a more detailed review and provide more detailed conclusions through a comparative table.
• Against the background of the emerging technologies mentioned above, we review the EMV-TPC strategies at intersections and EMV-TPC strategies on the road links simultaneously, which significantly guarantee EMV rapid traversal along the whole rescue route. Furthermore, we categorize them into several subclassifications respectively and provide helpful conclusions through a comparative table.

The organization of this paper is outlined as follows. Section 2 presents a bibliometric analysis of previous studies using VOSviewer software. Section 3 provides a review of EMV-TTP methods. Section 4 reviews the studies on EMV-RO algorithms. Section 5 summarizes the literature on EMV-TPC strategies. Section 6 outlines the potential avenues for future research. Section 7 provides the conclusion of this paper.

2. Review Methodology

2.1. Search Strategy

In this investigation, we adopted the systematic literature review methodology outlined by Booth et al. [20], with a specific focus on reviewing the existing research on routing and traffic priority control optimization for EMVs to identify the research gaps and potential future directions. We followed the three-stage process proposed by Tranfield et al. [21], which involved planning the review (Stage I), conducting the review (Stage II), and concluding the findings (Stage III).

Stage I focused on investigating the theoretical methodologies and practical applications related to the optimization of EMV routing and prioritization, with an emphasis on emerging technologies, such as the IoV, as depicted in Figure 2. All the research papers were quickly scanned to ensure that they focused specifically on EMVs rather than SVs. In Stage II, a comprehensive literature review was performed in March 2024, based on the Web of Science (WoS) Core Collection [22]. This online database offers extensive bibliographic information and analysis tools for assessing research trends and performance and includes high-quality journal articles and conference papers from prominent publishers such as ScienceDirect, IEEE, Spring, Nature, Sage, Wiley, Transportation Research Board National Research Council, Taylor Francis, et al. The review scope of this study began with the earliest year of data contained within the WoS Core Collection purchased by Tongji University, which was 1975, and provided sufficient temporal coverage for our research.

As listed in

Table 1. Search strings in this review.

Component	Search Strings in WoS
EMV-TTP	(TOPIC: “Emergency Vehicle*” OR “Rescue Vehicle” OR “Ambulance” OR “Fire Truck” OR “Police Car” OR “First Responder” OR “Privileged Vehicle” OR “Fire Engine”) AND (TOPIC: “Response Time” OR “Travel Time”) AND (TOPIC: “Estimat*” OR “Predict*” OR “Forecast*”)
EMV-RO EMV-TPC	(TOPIC: “Emergency Vehicle*” OR “Rescue Vehicle” OR “Ambulance” OR “Fire Truck” OR “Police Car” OR “First Responder” OR “Privileged Vehicle” OR “Fire Engine”) AND (TOPIC: “Path Optimization” OR “Route Planning” OR “Path Planning” OR “Traffic Signal” OR “Path Selection” OR “Fastest Path” OR “Shortest Path” OR “Route Optimization” OR “Pre-emption” OR “Pre-Emption” OR “Intelligent” OR “Connected” OR “Traffic Control” OR “Vehicle-to-Vehicle” OR “Connected Infrastructure” OR “Give Way” OR “Give the Way” OR “Lane*” OR “Exclusive Lane”)

*: Search result updated on 6 March 2024. TOPIC included the Title, Abstract, and Keywords.

, the research query used in the travel time prediction comprised three major components: “Emergency Vehicle”, “Response Time”, and “Prediction”, which were connected by an AND connector. The research query used in other parts of this review consisted of two main components: “Emergency Vehicle” and “Route Optimization”. These components were also linked by an AND connector. To expand the search scope and ensure the accuracy of the retrieved results, we included additional related terms in the research query, as listed in Table 1. Furthermore, to avoid irrelevant literature related to medical, legal, and other disciplines, we also added related terms such as “Traffic Signal” into the query to refine the retrieved results. Finally, we incorporated terms like “Intelligent”, “Connected”, and “Vehicle-to-Vehicle” into the query to keep up with emerging technologies development.

In Stage III, a complete full-text review was performed using a set of inclusion and exclusion criteria to guarantee the quality and relevance of the reviewed papers, as illustrated in

Table 2. The set of exclusion and inclusion criteria.

Inclusion Criteria	Exclusion Criteria
✓ English language papers ✓ Journal articles or conference papers ✓ Urban road network area ✓ Peer-reviewed papers	✓ Papers in languages other than English ✓ Book chapters, reports, letters, proceeding abstract ✓ Inter-city rescue ✓ Evacuation scenarios

. The literature search was restricted to academic journals and conference papers published in English, without imposing any restrictions on the publication year. To obtain more related research from the already retrieved papers, we utilized both forward and backward snowballing search strategies, following the methodology detailed by Jalali and Wohlin [23]. The backward snowballing search involved scanning related references included in the papers retrieved from WoS. Then, we conducted a forward literature search step using the “cited by” filter on Google Scholar to locate the recent contributions in this field. Finally, based on the papers retrieved from WoS, we conducted a bibliography analysis using VOSviewer (see Section 2.2).

2.2. Bibliography Analysis

In this subsection, we conduct the bibliometric analysis to examine the current status and explore developments in strategies for reducing EMV response times. Bibliometric analysis is a powerful quantitative tool used to derive statistical results for research performances and the contributions of journals, authors, countries, institutions, and more, as described by Kaffash et al. [24]. Based on the search results of Section 2.1, we employed VOSviewer to create the clustering networks shown in Figure 3 and Figure 4. VOSviewer is a software application developed by van Eck and Waltman in 2010 for visualizing bibliometric maps in bibliometric analyses. VOSviewer enabled us to display literature maps in various ways, each describing the literature from different perspectives [25].

Mapping the co-occurrence of keywords is a commonly used method to capture the association between co-occurring keywords. In this study, 3086 keywords appeared, and we selected the 99 keywords that occurred the most frequently (at least 8 occurrences) to construct the keyword co-occurrence network, as depicted in Figure 3. The size of each node signifies the frequency of each keyword’s appearance. A larger node indicates that the keyword occurred more often in the dataset. The length of the links connecting the nodes reflects the degree of relatedness among the keywords.

We divided these keywords into 5 clusters based on the automatic clustering result generated by the software.

Cluster 1 (green nodes) focused on emergency medical services, involving research related to “Survival”, “Ambulance”, “Cardiac Arrest”, “Cardiopulmonary Resuscitation”, and “Response Time”, reflecting the pivotal role of emergency medical services in enhancing the success rate of treatment and reducing mortality.

Cluster 2 (purple nodes) primarily focused on emergency response times and system operation statuses, including research related to “Time”, ”System”, and “Impact”, emphasizing the importance of reducing rescue response times to improve rescue success rates. It also reflects the application value of reducing negative traffic system impacts in enhancing urban emergency management capabilities and sustainability.

Cluster 3 (yellow nodes) focused on modeling optimization and algorithm development, including research related to “Model”, ”Algorithm”, ”Simulation”, and ”Traffic Control”. This highlights the importance of utilizing optimized models and algorithms to improve the efficiency of emergency responses.

Cluster 4 (red nodes) primarily focused on intelligent transportation systems and the Internet of Vehicles, covering emerging technologies such as “Artificial Intelligence” and “Autonomous Vehicles”. This cluster indicates the potential of intelligent transportation systems and Internet of Vehicles technologies in enhancing the efficiency of emergency vehicle responses and ensuring road safety.

Cluster 5 (blue nodes) mainly focused on the application of information technologies such as cloud computing, the Internet of Things (IoT), and GPS in the field of emergency rescue. This reflects the key role of information technology in ensuring emergency rescue communication and optimizing decision-making in emergency rescue services.

The network visualization of the co-citations of journals in the research on reducing response times was mapped, as depicted in Figure 4.

Cluster 1 (green nodes) pertained to journals in the field of emergency medicine and public health, focusing on top-tier journals in emergency medical services, public health, trauma care, and cardiovascular health. Journals such as “Resuscitation”, “Prehospital Emergency Care”, and “Accident Analysis & Prevention” had a high citation volume, indicating that researchers and professionals in the field of emergency vehicle rescue highly value rapid and effective emergency response.

Cluster 2 (blue nodes) was related to journals in the fields of transportation engineering and safety science, covering various aspects of traffic engineering, transportation planning, safety science, and traffic research. Journals such as “Transportation Research Record”, “Transportation Research Part B: Methodological”, and “Transportation Research Part C: Emerging Technologies” had high citation volumes, highlighting the importance of researching traffic safety, planning optimization, and the application of emerging technologies in emergency vehicle passage and the allocation of rescue resources.

Cluster 3 (yellow nodes) pertained to journals in the fields of operations research and transportation logistics, encompassing studies on operations research, soft computing, industrial engineering, and logistics management. Journals such as “European Journal of Operational Research” and “Computers & Operations Research” had high citation volumes, indicating that research on optimization models, decision support systems, and traffic logistics management plays a crucial role in improving the efficiency and management level of emergency vehicle rescue operations.

Cluster 4 (red nodes) was associated with journals in the fields of information technology and intelligent transportation systems, focusing on applied science, computer communications, intelligent transportation systems, the Internet of Things, and wireless communication technologies. Journals like “IEEE Transactions on Intelligent Transportation Systems”, “IEEE Access”, and “IEEE Transactions on Vehicular Technology” had high citation volumes, underscoring the significant role of information technology in enhancing the speed of emergency service responses, improving road safety, and increasing traffic management efficiency.

These findings underscore that reducing response times is an interdisciplinary issue in academic research.

3. EMV-TTP Methods

Precise EMV-TTP methods serve as the foundation for achieving optimal EMV routing outcomes. Compared with SVs, EMVs have unique driving policies and possess distinct physical attributes, including larger vehicle width, a higher center of gravity, and a larger turning radius [26]. Therefore, EMVs’ actual historical travel data are the most reliable data source for EMV-TTP [27]. However, the limited availability of such data resulted in constrained prediction accuracy and practicability in the existing research. Based on the specific process for predicting the travel time of emergency vehicles, we divided this into three stages: data collection, method selection, and result application, as shown in Figure 5, illustrated by real-world scenarios.

Considering the differences and similarities among previous studies, we classified the studies into four categories: (1) model-based methods, (2) simulation-based methods, (3) API-based methods, (4) UAV-based methods.

3.1. Model-Based Methods

Relevant studies on model-based methods considered EMVs’ travel times in a specific time period as either a constant or a random value. Model-based methods were primarily used for emergency infrastructure site planning and EMV dispatching [28]. While these methods can reduce model complexity, they cannot fully capture the vehicle dynamics and features of EMVs and time-dependent features of road traffic flows. To address these limitations, some studies modeled the vehicle dynamics of EMVs and the variance of traffic conditions by proposing comprehensive formulas, including a segmentation function-based model [29], semi-parametric prediction model [30], regression models [31,32], Bureau of Public Road function (BPR function) [33,34,35], and Bayesian neural network [36].

3.2. Simulation-Based Methods

Simulation-based methods were developed using traffic simulation tools (Cellular Automata (CA), SUMO) to predict EMV travel times [37,38]. Simulation-based methods can simulate the specific driving behavior of EMVs (i.e., frequent lane-changing and running red lights) [37] and their interaction behaviors with SVs (i.e., emergency message (EM) broadcast) [38]. Using SUMO software, Agarwal and Paruchuri [38] developed a vehicle–vehicle communication-based model for EMV driving and lane change behavior on road segments. Małecki et al. [39] implemented a multi-cell CA model for a road segment with two lanes in one direction to discover the lateral shift behaviors of EMVs and frictional conflicts when meeting SVs. The simulation results suggested that road width, shoulder width, and driver behavior during EMV passage significantly impact EMV travel times. In summary, traffic simulation tools could effectively integrate EMV-TPC strategies and the actual driving behavior of EMVs and SVs into EMV-TTP.

3.3. API-Based Methods

With the growth of information technology, several studies utilized map APIs (e.g., Google API, AutoNavi API) to acquire real-time traffic conditions [40]. These APIs gather anonymous location data from map users (most users are SVs) and traffic cameras to conduct EMV-TTP. Hence, the accuracy of API data is highly dependent on the data source. Furthermore, due to the difference between SVs and EMVs, the prediction result cannot be used for EMVs directly. Fleischman et al. [41] developed a linear regression model to estimate the gap between ambulance historical travel times and travel times obtained from Google API in Oregon, USA. The results showed that travel time estimation based solely on the Google APIs led to significant underestimation. Torres et al. [42] utilized a machine learning approach to predict the gap mentioned above, improving the accuracy of the API-based methods for EMV-TTP.

3.4. UAV-Based Methods

Additionally, several studies explored novel approaches to monitor traffic conditions, using UAVs to improve EMV-TTP accuracy during abnormal scenarios such as earthquakes [43,44,45,46]. During natural disasters such as floods or earthquakes, traffic facilities like detectors, geomagnetic coils, and other roadside units can be damaged by power outages, making it difficult to detect congestion or accidents along the rescue route [47]. Under these scenarios, UAVs are immune to road disruptions and traffic congestion and aid in monitoring emergency events and providing real-time feedback about traffic conditions beyond direct line of sight (e.g., congestion, accidents), helping to reduce response times [43,48]. Oubbati et al. [49] estimated the road segment travel times for EMVs through a macroscopic model incorporating traffic data from UAVs. Beg et al. [50] proposed an autonomous traffic policing system that detects real-time emergency traffic situations and potential accidents using UAV networks.

3.5. Discussion

This section discusses several topics related to EMV-TTP. Some discussions for future research are presented below. Firstly, model-based methods have well-structured frameworks with interpretable parameters [27]. They are computationally efficient and enable effective scaling of extensive road networks and historical trip databases [27]. However, they still have gaps from reality, such as the ignoration of signal control in the BPR function. Secondly, future research should focus on using actual EMV data (e.g., trajectory data) and API data to improve the practicability of EMV-TTP methods. Finally, researchers should account for factors such as sudden disruptions in traffic systems during various types of disasters. Sophisticated tools such as UAVs and satellites can be utilized for the 24/7 monitoring of rescue areas to ensure the precise and up-to-date monitoring of changes in traffic conditions.

4. EMV-RO Algorithms

Similar to SVs, EMVs usually select the best route between the emergency station and rescue spot to maximize their utilities. However, in addition to the differences in vehicle size, vehicle dynamics, and driving policies mentioned before, EMVs also differ from SVs as EMVs are more concerned with travel times [14], safety risks [51], and reliabilities [52], rather than comfort, fuel consumption, and travel distance in route optimization objectives. To better understand this process, we divided it into three stages: environmental perception, feature extraction, and algorithm selection, as illustrated in Figure 6, accompanied by real-world scenarios.

Considering the differences and similarities among previous studies, we categorized EMV-RO into three categories: EMV static routing optimization (EMV-SRO), EMV dynamic routing optimization (EMV-DRO), and EMV routing optimization with additional support (EMV-ROAS).

4.1. EMV-SRO Algorithms

EMV-SRO aimed to find the shortest or fastest routes for EMVs on a static road network. The cost function of each edge comprised one or more constant parameters, such as road length [53], travel time [13], speed limit [54], and occupancy [55], which remained unchanged throughout the study time period, as shown in Table 3. For instance, Nordin et al. [53] developed an interface using C# based on the A* algorithm to help ambulances in Malaysia find the best routes with the shortest distance. Zhao et al. [13] developed a multi-objective optimization (MOO) model that considered travel time, travel reliability, and travel safety in the objective function when rescuing on a network with reduced road capacities and urban road congestion due to sudden disasters. Moreover, Constantinescu and Pătrașcu [54] presented a genetic algorithm to solve the multi-objective route planning optimization problem by considering road congestion and speed limit. Al Mustafid [56] proposed a fuzzy Dijkstra algorithm for finding the shortest route for EMVs, where the cost function associated with each road segment was determined by considering both the density of the road and its length.

In addition, some relevant studies segmented a day into various time periods, including morning rush hours, evening rush hours, and periods outside of peak times [57,58]. These studies considered the road condition variations among different time periods but treated the road condition within each time period as constant. Thus, these studies [57,58] could still be grouped into the EMV-SRO algorithm. Moreover, compared with some heuristics algorithms (e.g., Genetic Algorithm [54], A* Algorithm [53]), deterministic algorithms, such as Dijkstra [55] and its variant [57] are more effective in solving EMV-SRO algorithms by providing accurate results while requiring less computation time and exhibiting good scalability [59]. However, EMV-SRO algorithms cannot adapt to dynamic urban road networks with constantly changing environments (e.g., road congestion). Therefore, it is crucial for EMVs to dynamically refine their rescue route based on real-time traffic conditions, which promotes the development of the EMV-DRO algorithms.

4.2. EMV-DRO Algorithms

Basically, the most popular approach for addressing EMV-DRO algorithms is through the online re-planning of the remaining route, from the updated location of an EMV to the rescue spot, when the traffic condition of the road network changes. Usually, this online re-planning of the remaining route is considered as a sequence of EMV-SRO problems. This is because online re-planning treats the road network as static when adjusting the remaining route. Consequently, we referred to similar methods such as Emergency Vehicle Online Route Re-Planning (EMV-ORRP).

Regarding EMV-ORRP, Fu et al. [60] introduced a chance-constrained programming model solved by the Particle Swarm Optimization (PSO) algorithm. This model considered the situation where multiple emergency response departments sent EMVs to a single rescue spot, and the travel times between any pair of nodes were characterized as stochastic variables. Similar to Fu et al. [60], Wang et al. [61] also utilized the PSO algorithm to find the fastest routes for EMVs, while Wang et al. [61] incorporated the travel times of EMVs passing through the intersections in their optimization model, which reduced the inaccuracies caused by ignoring intersection passing time for EMVs in previous studies. To provide dynamic route guidance for EMVs, Chen et al. [62] proposed a modified Dijkstra algorithm and combined the Starlogo software with MATLAB to simulate the guidance process. Amer et al. [63] introduced an algorithm called the Coalition Game Approach based on PSO (CGA-PSO), aimed at reducing EMV rescue times, fuel consumption, and emissions and increasing average speeds during the rescue process. Wang and Zlatanova [64] proposed an improved Dijkstra algorithm to generate a feasible and safe route for first responders to minimize the travel risk while constraining the rescue travel time. It also considered the impacts of different types of disasters on traffic networks. Oubbati et al. [44] suggested a UAV-based method to oversee the status of road segments and occurrences (including road length, total count of vehicles, average velocity, and the degree of vehicle distribution). It provided the optimal route grounded on the Dijkstra algorithm for EMVs with the shortest arrival times. To reduce EMV travel times, Duan et al. [65] developed a two-stage optimization model that consisted of the Dijkstra algorithm and Salp Swarm Algorithm (SSA) for finding alternative routes (Stage 1) followed by an improved SSA with a population grouping strategy for obtaining the optimal EMV rescue route and SV evacuation schemes (Stage 2). Jose and Vijula Grace [2] devised a dynamic routing strategy for EMVs that consisted of two main elements: the discovery of K-routes and the selection of the optimal route, employing the Exponential Bird Swarm (Exp-BSA) algorithm for this purpose. The Exp-BSA algorithm outperformed existing algorithms in various aspects, including travel distance, average traffic density, average speed, and the travel time associated with the rescue route. Mahariba et al. [11] proposed a multiple-source shortest travel time (MSSTT) algorithm combined with GIS software to provide optimal routes for ambulances in developing countries. Liu et al. [35] introduced a cloud-based and service-oriented cooperative route planning mechanism for heterogeneous vehicles, i.e., SVs and EMVs. This mechanism leveraged an A*-based gradual path planning algorithm to offer varying priority levels of route planning services for all vehicles within the road network. Heda et al. [66] provided an optimal rescue route for ambulances, considering the real-time traffic status (travel time and traffic congestion) from the TomTom Routing API and TomTom Traffic API. Therefore, overall, most studies selected travel time as the cost function of each edge [35,60,62]. Some research also selected other variable parameters, including average speed [44], traffic jam index [65], total vehicle number [44], and density [63], as the cost function. Various heuristics algorithms, including the PSO algorithm [60,61] and its variant [63] and the A* algorithm and its variant [35], have been proposed to solve the EMV-ORRP to improve the solving efficiency, compared with the Dijkstra algorithm and its variants [62,64].

Apart from EMV-ORRP, Hu et al. [59] first introduced a method called co-evolutionary path optimization (CEPO), aimed at addressing path optimization within a specified dynamic routing environment by predicting areas on the road network that may be obstructed by moving obstacles. The CEPO method achieved higher computing efficiency than EMV-ORRP by utilizing the ripple-spreading algorithm (RSA), which combined the multi-step updates into one step. However, their model focused on SVs rather than EMVs. Building upon the work of Hu et al. [59], Wen et al. [67] further developed a methodology termed timing co-evolutionary path optimization (TCEPO), which combined the CEPO and EMV-ORRP to optimize EMV routes.

4.3. EMV-ROAS Algorithms

EMV-DRO has shown significant advancements over EMV-SRO, even when not combined with EMV-TPC strategies at intersections and on road links [68]. Further, considering the differences regarding driving policies between various EMVs and SVs, researchers began to focus on EMV-ROAS algorithms to reduce EMV response times by combining EMV-SRO/DRO with additional support from traffic departments, including EMV-TPC at intersections [68,69], lane reservation on road links [70,71], and emergency broadcast [72] to minimize EMV response times. Gedawy [68] combined a dynamic route planning algorithm based on the D* Lite informed search algorithm with signal priority control to reduce the travel times of EMVs further. Maximizing other traffic flow through the intersection was taken as a secondary optimization objective. Shaaban et al. [72] identified the optimal route for an EMV using the Dijkstra algorithm and applied traffic signal priority control at every intersection along the chosen path. After the optimal route was confirmed, an emergency message was transferred to other SVs via the IoV. Nguyen et al. [70] presented a macroscopic path planning framework, incorporating the k-shortest route planning algorithm, lane reservation, and traffic signal priority control. This framework was designed to reduce the travel times of EMVs. Su et al. [71] proposed a decentralized Reinforcement Learning (RL) framework called EMVLight, which combined dynamic EMV routing and traffic signal priority control to minimize the travel times for both EMVs and SVs. Dutta et al. [73] introduced a strategy that integrated vehicle rerouting and traffic signal optimization at intersections, modeling the urban dynamic traffic environment with priority vehicle passage requirements as a bilevel optimization problem. The upper-level objective aimed to minimize the average travel time of all vehicles while giving priority to EMVs. The lower-level objective focused on achieving network equilibrium. The details of additional support will be described in Section 5. It is noteworthy that studies in terms of the EMV-ROAS were not only concerned with EMV response times but also with the optimization objectives of the SVs, such as SV travel times [72] or SV maximum throughput [68].

4.4. Discussion

In this section, we discuss the EMV-RO studies in three categories. Based on the comparison results (Table 3), some discussions for future research are presented below. Firstly, most studies optimized EMV routes based on travel distance and travel time, but only a few studies incorporated factors, including occupancy, traffic density, and risk indices, in the objectives. Secondly, numerous studies utilized the Dijkstra and A* algorithms and their variants as the solving algorithms. With advancements in information technology, some researchers started to utilize intelligent algorithms, including PSO and RSA. Hence, it is crucial to develop suitable algorithms with high accuracy and efficiency for EMV real-time routing in the future. Thirdly, most studies solely focused on optimizing EMV routing, and few studies in terms of EMV-RO paid attention to the benefits of various EMV-TPC strategies simultaneously, which could improve the rescue efficiency and safety of EMVs. Fourthly, future research should account for the significant differences between EMVs and SVs, including optimization objectives, vehicle attributes, vehicle dynamics, driving policies, and driver preferences.

Table 3. Analysis of the EMV routing algorithms.

Author (Year)	Categories	Solving Algorithm	Optimization Objectives	Cost Functions	R.O.
Derekenaris et al. (2001) [55]	EMV-SRO	DA	EMV RT	L, SL, O	A
Nordin et al. (2012) [53]	EMV-SRO	A Star Algorithm	EMV TD	L	A
Musolino et al. (2013) [57]	EMV-SRO	Modified DA	EMV RT	T	G
Zhao et al. (2017) [58]	EMV-SRO	KSPA	EMV RT	T	G
Zhao et al. (2019) [13]	EMV-SRO	MOO Algorithm	EMV RT, DS, reliability	T	G
Constantinescu and Pătrașcu (2020) [54]	EMV-SRO	Genetic Algorithm	EMV RT	L, SL, O	G
Al Mustafid et al.(2022) [56]	EMV-SRO	Fuzzy DA	EMV TD	L	G
Hussein et al. (2022) [74]	EMV-SRO	Neural Network	EMV TD	L	A
Fu et al. (2010) [60]	EMV-DRO	PSO Algorithm	EMV RT	T	G
C. Wang et al. (2013) [61]	EMV-DRO	PSO Algorithm	EMV RT	T	G
Chen et al. (2014) [62]	EMV-DRO	Improved DA	EMV RT	T	G
Amer et al. (2018) [63]	EMV-DRO	Adjusted PSO Algorithm	EMV RT, emissions, et al.	L, D, T	G
Wen et al. (2020) [75]	EMV-DRO	Adjusted RSA Algorithm	EMV RT	T	G
Wang and Zlatanova (2020) [64]	EMV-DRO	Modified DA	EMV RT, DS	T, risk index	G
Oubbati et al. (2021) [44]	EMV-DRO	DA	EMV RT	L, AS, total vehicle number	A
Duan et al. (2022) [65]	EMV-DRO	DA (with SSA)	EMV RT	L, T, traffic jam index	G
Jose and Vijula Grace (2022) [2]	EMV-DRO	KSPA (with Exp-BSA)	EMV RT, TD, et al.	L, D, T, AS	G
Mahariba et al. (2022) [11]	EMV-DRO	MSSTT Algorithm (with GIS)	EMV RT	T	G
Liu et al. (2023) [35]	EMV-DRO	Gradual Path Planning Algorithm	EMV RT, TE	T	G
Gedawy (2008) [68]	EMV-ROAS	D Star Lite Informed Search Algorithm	EMV RT, SV MT	T	G
Shaaban et al. (2019) [72]	EMV-ROAS	DA	EMV RT, SV TT	T	G
Nguyen et al. (2022) [70]	EMV-ROAS	KSPA	EMV RT	L, SL, number of lanes	G
Su et al. (2023) [71]	EMV-ROAS	Decentralized DA	EMV RT, SV TT	T	G

Note 1: In column Solving Algorithm: DA for Dijkstra algorithm, KSPA for k-shortest path algorithm. Note 2: In column Optimization Objectives: RT for response time, TT for travel time, TD for travel distance, DS for driving risk, MT for maximum throughput, and TE for traffic equilibrium. Note 3: In column Cost Functions: T for travel time, L for road length, O for occupancy, D for density, AS for average speed, and SL for speed limit. Note 4: In column Research Objective (abbreviated as R.O.): G for general EMV, A for ambulance.

Table 3. Analysis of the EMV routing algorithms.

Author (Year)	Categories	Solving Algorithm	Optimization Objectives	Cost Functions	R.O.
Derekenaris et al. (2001) [55]	EMV-SRO	DA	EMV RT	L, SL, O	A
Nordin et al. (2012) [53]	EMV-SRO	A Star Algorithm	EMV TD	L	A
Musolino et al. (2013) [57]	EMV-SRO	Modified DA	EMV RT	T	G
Zhao et al. (2017) [58]	EMV-SRO	KSPA	EMV RT	T	G
Zhao et al. (2019) [13]	EMV-SRO	MOO Algorithm	EMV RT, DS, reliability	T	G
Constantinescu and Pătrașcu (2020) [54]	EMV-SRO	Genetic Algorithm	EMV RT	L, SL, O	G
Al Mustafid et al.(2022) [56]	EMV-SRO	Fuzzy DA	EMV TD	L	G
Hussein et al. (2022) [74]	EMV-SRO	Neural Network	EMV TD	L	A
Fu et al. (2010) [60]	EMV-DRO	PSO Algorithm	EMV RT	T	G
C. Wang et al. (2013) [61]	EMV-DRO	PSO Algorithm	EMV RT	T	G
Chen et al. (2014) [62]	EMV-DRO	Improved DA	EMV RT	T	G
Amer et al. (2018) [63]	EMV-DRO	Adjusted PSO Algorithm	EMV RT, emissions, et al.	L, D, T	G
Wen et al. (2020) [75]	EMV-DRO	Adjusted RSA Algorithm	EMV RT	T	G
Wang and Zlatanova (2020) [64]	EMV-DRO	Modified DA	EMV RT, DS	T, risk index	G
Oubbati et al. (2021) [44]	EMV-DRO	DA	EMV RT	L, AS, total vehicle number	A
Duan et al. (2022) [65]	EMV-DRO	DA (with SSA)	EMV RT	L, T, traffic jam index	G
Jose and Vijula Grace (2022) [2]	EMV-DRO	KSPA (with Exp-BSA)	EMV RT, TD, et al.	L, D, T, AS	G
Mahariba et al. (2022) [11]	EMV-DRO	MSSTT Algorithm (with GIS)	EMV RT	T	G
Liu et al. (2023) [35]	EMV-DRO	Gradual Path Planning Algorithm	EMV RT, TE	T	G
Gedawy (2008) [68]	EMV-ROAS	D Star Lite Informed Search Algorithm	EMV RT, SV MT	T	G
Shaaban et al. (2019) [72]	EMV-ROAS	DA	EMV RT, SV TT	T	G
Nguyen et al. (2022) [70]	EMV-ROAS	KSPA	EMV RT	L, SL, number of lanes	G
Su et al. (2023) [71]	EMV-ROAS	Decentralized DA	EMV RT, SV TT	T	G

Note 1: In column Solving Algorithm: DA for Dijkstra algorithm, KSPA for k-shortest path algorithm. Note 2: In column Optimization Objectives: RT for response time, TT for travel time, TD for travel distance, DS for driving risk, MT for maximum throughput, and TE for traffic equilibrium. Note 3: In column Cost Functions: T for travel time, L for road length, O for occupancy, D for density, AS for average speed, and SL for speed limit. Note 4: In column Research Objective (abbreviated as R.O.): G for general EMV, A for ambulance.

5. EMV-TPC Strategies

Although EMV-RO can reduce response times, EMVs are still susceptible to traffic lights and traffic congestion due to the highly dynamic time-dependent features of the urban traffic system [76]. Therefore, the traffic department should implement EMV-TPC strategies while EMVs perform rescue tasks. Based on the existing studies and implications, we divided these strategies into two main components, as reviewed in Section 5.1 and Section 5.2.

5.1. EMV-TPC at Intersections

In recent decades, numerous studies have focused on developing EMV-TPC systems at intersections [77,78,79]. Similar to freight and transit signal priority, the process of EMV-TPC strategies at intersections can be divided into four stages: Stage 1: Normal; Stage 2: Detection; Stage 3: Priority Control; Stage 4: Transition, as shown in Figure 7.

5.1.1. Stage 1: Normal

In Stage 1, intersections are generally equipped with fixed-time traffic control systems that operate consistently throughout the day or during specific periods. In other cases, some intersections may employ adaptive traffic control systems, dynamically adjusting the signal scheme based on the time-dependent traffic flow in different directions.

5.1.2. Stage 2: Detection

In Stage 2, EMVPS commonly employ light/infrared-based [80], sound-based [81], fixed sensors-based [82], radio-based [76,83], GPS-based [84,85], and VANET-based [71,86] solutions to detect the EMV-TPC demand within the control area, as shown in Figure 5. Upon detection of the entry of an EMV or receiving a request from an EMV, the traffic signal controller (TSC) can be quickly activated to remain or turn green upon confirmation of the request.

Each solution discussed above has its advantages and disadvantages. It is essential to leverage their strengths to their full potential in practice. However, implementing them in the real world poses several challenges, such as false triggering, communication delay, obstruction of view, and high costs [76]. As a result, researchers should consider these issues during the simulation process to minimize the gap between simulation results and real-world applications.

5.1.3. Stage 3: Priority Control

In Stage 3, the EMV-TPC system responds to requests from EMVs by adjusting the TSC from Stage 1, Stage 2, and Stage 3 to remain or turn green. Based on existing studies, we categorized studies into three types: node-based optimization (one single intersection), path-based optimization (neighboring intersections along the route), and network-based optimization (multiple intersections on the road network).

Node-based optimization focuses more on the modification mechanism of the TSC and the interactions between EMVs and TSC [76,82,83,85]. Jagadeesan et al. [85] established a control system consisting of RFID, GPS, and Long Term Evolution (LTE) to assist ambulance movement. GPS played a key role in triggering changes in traffic signal modes as an ambulance neared by supplying information on speed, distance, and projected time of arrival. RFID technology aided in verifying an ambulance’s transit past the traffic signals. The proposed system was tested with the AnyLogic software. Rosayyan et al. [76] proposed a methodology combined with a decentralized RF and Global Navigation Satellite System (GNSS) based on the geo-fencing approach to help EMVs pass intersections with minimum delays. The proposed method was field-tested at an intersection in India and evaluated by an external agency. To enhance the performance of EMV-TPC strategies, researchers are exploring more sophisticated techniques aimed at obtaining precise spatial–temporal information on EMVs such as RFID [85] and high-precision GPS [87]. However, given that EMV rescue routes frequently traverse a number of signalized intersections, node-based optimization methods would disrupt regular traffic flow patterns for each intersection sequentially, which inevitably entails significant adverse effects on SVs.

Therefore, some researchers focused on path-based optimization, which provides route guidance and a “green wave” (i.e., signal coordination of neighboring intersections) along the rescue route [1,80,88]. Kang et al. (2014) [89] proposed an EV signal coordination (EVSC) approach that provided a “green wave” for EVs. Traffic simulations were conducted along an emergency corridor with eight intersections in Qingdao, China. Using two-level programming, Yao et al. [1] developed an EMV signal-coordinated control model along rescue routes to minimize delays, travel times, queue lengths, and the number of stops for EMVs. The proposed model was simulated in three intersections in Suzhou, China, using the micro-traffic simulation software VISSIM. Mu et al. [90] introduced a signal priority control approach that utilized route information from EMVs to ascertain the optimal time range of green light intervals at each intersection. This method was designed with dual goals: to minimize the waiting time of EMVs at all intersections and to maximize the passing numbers of general SVs. Compared to node-based optimization, path-based optimization can reduce EMV travel times with fewer negative impacts on SVs. However, it should be noted that path-based optimization methods do not consider the TSCs at other intersections beyond the selected rescue route, which constrains the solution space.

Consequently, network-based optimization methods have been proposed that focus more on coordinating multiple intersections at the network scale to additionally mitigate the adverse effects on the road network resulting from prioritizing EMVs [71,79,86,91]. Petrică et al. [91] proposed a traffic light priority control system for EMVs and public transportation that reduced average emergency response times and optimized traffic flow at intersections. The proposed system was evaluated on the road networks of San Francisco, Rome, and Beijing using the Sim2Car simulator. Zhong and Chen [79] introduced a real-time traffic signal control strategy for EMVs that considered three critical indicators: the level of emergency response required, the congestion level on the road link, and the urgency level of the rescue time. The strategy was tested on a road network in Shanghai using the SUMO simulator. Cao et al. [86] delivered an early attempt to control traffic signals for EMVs by utilizing the Deep Q-Network algorithm. This method was tested within a traffic road network simulated by the SUMO software. Su et al. [71] introduced a decentralized RL framework designed for the dual purpose of dynamically routing EMV and controlling traffic signal priorities. The methodology allowed EMVLight to master cooperative traffic signal phasing strategies at the network level, which not only cut down on the travel times of EMVs but also shortened the travel times of SVs. Due to the limitations of real-world experiments, the verification of these algorithms was conducted in simulation software, such as SUMO [71,86]. The simulation results showed that the implementation of network-based optimization methods has the potential to minimize negative impacts on SVs throughout a whole road network. However, full simulation environments may deviate from reality to a certain extent, and, thus, on-site testing and verification are still required to validate the proposed strategies.

5.1.4. Stage 4: Transition

In Stage 4, once EMVs pass through intersections, it is essential that the TSCs at previous intersections transition from Stage 3 back to Stage 1 safely and expeditiously. This process may range from zero to five cycles, depending on the transition algorithm and cycle length [92]. However, limited studies proposed the transition methods after EMV-TPC operations [88,93], including the smooth transition algorithm [92], optimal control algorithm [78], MOO [88], and linear programming [79]. These algorithms aimed to minimize or maximize one or more of the optimization objectives, including minimizing transition time [79,92], minimizing queue length [78], and maximizing SVs passing in per unit time during Stage 4 [88]. In summary, diverse transition algorithms exhibit varied performances for intersections after the passage of EMVs. Consequently, forthcoming studies should select appropriate transition algorithms to minimize both transition times and negative impacts on SVs.

5.2. EMV-TPC on Road Links

Although EMVs benefit from the EMV-TPC at intersections, their rescue efficiencies can be negatively impacted by traffic congestion and the associated SVs when traveling on road links. In recent years, some studies have also proposed EMV-TPC strategies for urban road links. These include allowing EMVs to prompt SVs to give way to EMVs through traditional sirens and lights, engaging in vehicle-to-vehicle communication, establishing temporary or permanent EMV lanes, and utilizing shoulder lanes or opposite-direction driving, among other methods. To more clearly demonstrate this process, we drew a diagram framework for EMV-TPC strategies for road links, as illustrated in Figure 8, accompanied by real-world scenarios.

Considering the differences and similarities among the previous studies, we categorized these strategies into three groups: giving way to EMVs, setting up EMV lanes, and other strategies, as described below.

5.2.1. Giving Way to EMVs

Giving way to EMVs is a crucial responsibility that all SVs must uphold and adhere to on road links. When an EMV activates its L&S equipment, the associated SVs should immediately slow down and pull over to the side of the road to allow the approaching EMVs to pass. It is essential for EMVs and SVs to remain vigilant of their surroundings and ensure that it is safe to yield before doing so. However, despite being the most commonly used warning system of EMVs, the L&S system is only effective over short ranges and at low speeds [94].

In recent years, several studies have adopted VANET to analyze the various behaviors of SVs when giving way to EMVs in simulation environments [38,95,96,97]. Zhao et al. [97] analyzed the characteristics of SVs giving way to EMVs, considering the lane-changing behavior of EMVs, and developed a two-way, two-lane CA model to improve the passing speed of EMVs. Agarwal and Paruchuri [38] developed a lane-changing model for EMVs to simulate their driving behaviors on road links in SUMO. In this model, EMVs alerted SVs to give way through emergency messages based on the IoV. Wu et al. [96] introduced an EMV lane pre-clearing strategy to prioritize EMVs on road links through cooperative driving with surrounding connected SVs. Formulated by a mixed-integer nonlinear programming model, the strategy divided the SVs in front of the EMVs into several blocks and designed the optimal merging trajectories of connected SVs to ensure the desired speed of EMVs while minimizing the negative impact on connected SVs. Based on V2V communications, Osman et al. [95] devised a strategy to urge all SVs in front of the EMVs to leave the lane after receiving emergency messages. Based on the emergency messages received, the SVs adapted their speed and the maximum distance between any two vehicles to avoid collisions and accidents. Lin et al. [98] proposed a collaborative optimization model for multiple EMVs aimed at planning their driving trajectories on road links while not disturbing the normal operation of associated SVs. The goal was to maximize EMV traffic efficiency and find a balance between minimizing the negative impact on SVs and ensuring the smoothness of EMV driving trajectories. Cortés and Stefoni [99] analyzed EMV trajectory data and onboard video data from Santiago, Chile, to extract the interbehavior characteristics between EMVs and surrounding SVs. They calibrated parameters such as the average response time for SVs to give way to EMVs under various scenarios (changing lanes in front of EMVs, pulling over to the roadside or pedestrian side, and stopping near the intersection stop line to allow EMVs to pass) in an effort to overcome the fixed simulation threshold limitations of most commercial software (such as SUMO, VISSIM, etc.).

5.2.2. Setting up EMV Lanes

In scenarios where SVs give way to EMVs, some of the SVs in front of the approaching EMVs need to change lanes, rather than having all the SVs vacate the entire lane for EMVs. This strategy is easy to implement in a fully connected environment (e.g., SUMO, VISSIM). However, in current realistic scenarios, there are often a number of constraints, including obstructed dissemination of warning L&S and the inability of SV drivers to be informed of the exact yielding requirements (i.e., when to give way and which lane to leave) in a timely manner. As a result, several studies have focused on the early clearance of specific lanes to ensure that the expected speed of EMVs on certain road links can be maintained [71,81,100].

Under V2V communications, Cetin and Jordan [101] presented a strategy to split the vehicular queue on one of the two-lane road links at oversaturated intersections based on shockwave theory. The proposed strategy was simulated in the VISSIM. Patel et al. [81] proposed a framework that identified ambulances based on a neural network and helped set up a makeshift EMV-dedicated lane on road links along the rescue route. Su et al. [71] calculated the emergency capacity of the road link by the segment’s shoulder width, lane width, and other factors in order to determine whether an EMV lane could be set up or not. If there was insufficient capacity to set up an EMV lane, the speed of EMVs would be equal to that of the average speed of the SVs on this particular road link. Otherwise, the speed of the EMVs would achieve the maximum speed allowed for EMVs on this road link. Another option to consider would be exclusive transit lanes or high occupancy vehicle (HOV) lanes. Xie et al. [100] proposed an optimization model for designing and managing a multi-functional exclusive lane (MFEL). Implementing MFELs in a reasonable manner can reduce EMV response times and transit passengers’ travel times simultaneously. In practical applications, Agrawal and Maheshwari [102] proposed a smart, movable road divider system to automatically recognize an ambulance and clear a lane using this device. Additionally, lane control signs could also be employed as a solution to set up an EMV lane in advance.

5.2.3. Other Strategies

In addition to giving way and setting up EMV lanes, several studies have explored other EMV-TPC strategies for road links under emergency circumstances. Due to the unique characteristics of EMVs that are not restricted by driving direction, Zhao et al. [103] established a two-way, two-lane cellular automaton model that allowed EMVs to use opposite lanes for overtaking. While the simulation results showed that this approach could effectively increase the average speed of EMVs, it may also interfere with the normal driving of SVs traveling in the opposite lane. Wang et al. [104] included opposite lanes in the design of urban road traffic networks and formulated a bilevel optimization model to optimize the design of EMV lanes and the routing of SVs. Numerical experiments demonstrated benefits for both EMVs and SVs resulting from the implementation of EMV lanes.

Another strategy is the utilization of shoulder lanes. Under normal circumstances, shoulder lanes can only be used by cyclists and motorcyclists. When EMVs enter the communication range, SVs driving in adjacent lanes should move into the shoulder lane to form an EMV lane ahead of the approaching EMVs to reduce the delay of EMVs [84]. Additionally, Wang et al. [105] conducted a study on road closure control for EMVs when the degree of priority achieved the highest value, which meant that the efficiency and safety of EMVs were guaranteed to the highest degree, but the negative impact on the surrounding road network was significant.

However, these strategies have non-negligible driving risks without additional support from the traffic department.

5.3. Discussion

In this section, we discuss EMV-TPC strategies for intersections and road links. According to the comparison results (

Table 4. Analysis of traffic priority strategies for EMVs.

Author (Year)	Control Range	Detection	Priority Techniques	Negative Impact	T.A.	C.D.	R.O.
Unibaso et al. (2010) [83]	Node	Radio, GPS	SPC	No	No	No	G
Weng et al. (2011) [82]	Node	Fixed Sensors	SPC	No	Yes	No	G
Li et al. (2018) [107]	Node	VANET (Field Test)	SPC	No	No	Yes	G
Jagadeesan et al. (2019) [85]	Node	RFID, GPS, LTE	SPC	No	No	No	A
Mu et al. (2020) [88]	Node	VANET (Numerical)	SPC	SV PC	Yes	No	G
Rosayyan et al. (2022) [108]	Node	VANET (Matlab), EC	SPC	No	No	Yes	G
Humayun et al. (2022) [84]	Node	5G, GPS, IoT, CC	SPC, SL	SV TT	No	No	G
Rosayyan et al. (2023) [87]	Node	GPS, IoT, EC	SPC	SV Waiting Time	No	Yes	G
Zhao et al. (2015) [103]	Link	Numerical Simulation	Opposite Lane	No	No	No	G
Agarwal and Paruchuri (2016) [38]	Link	VANET (SUMO)	GWE	SV DT	No	No	G
Wu et al. (2020) [96]	Link	VANET (Numerical)	GWE	No	No	No	G
Osman et al. (2021) [95]	Link	VANET (Matlab, NS-2)	GWE	No	No	Yes	G
Alzubaidi et al. (2023) [109]	Link	VANET (Numerical)	GWE	No	No	No	G
Shibuya et al. (2000) [80]	Path	Infrared Beacon	SPC	No	No	No	G
Nelson and Bullock (2000) [92]	Path	Not Mentioned	SPC	S’ TT	Yes	No	G
Qin and Khan (2012) [78]	Path	VANET (Matlab)	SPC	SV Queue Length	Yes	No	G
Wang et al. (2013a) [105]	Path	VANET (CA)	GW, EMV Lane et al.	SV TT and DT	No	No	G
Yang et al. (2014) [110]	Path	VANET (HIL)	SPC	SV DT	No	No	G
He et al. (2014) [111]	Path	VANET (VISSIM)	SPC	SV DT et al.	No	No	G
Yao et al. (2018) [1]	Path	VANET (VISSIM)	GW	SV DT	No	No	G
Mu et al. (2018) [90]	Path	Fixed Sensors	SPC	SV PC	No	No	G
Patel et al. (2022) [81]	Path	Acoustics, IoT, GPS	EMV Lane	No	No	No	G
Petrică et al. (2021) [91]	Network	VANET (Sim2Car)	SPC	SV TT et al.	No	No	G
Zhong and Chen (2022) [79]	Network	VANET (SUMO)	SPC	SV Waiting Time	Yes	No	G
Cao et al. (2022) [86]	Network	VANET (SUMO)	SPC	SV Queue Length et al.	No	No	G
Su et al. (2023) [71]	Network	VANET (SUMO)	SPC, EMV Lane	SV Average TT	No	No	G

Note 1: In table header: T.A. for transition algorithm, C.D. for communication delay, R.O. for research objective. Note 2: In column Detection: EC for edge computing, CC for cloud computing. Note 3: In column Priority Techniques: SPC for signal priority control, GW for green wave, GWE for giving way to EMVs, SL for shoulder lane. Note 4: In column Negative Impact: PC for passing capacity, TT for travel time, DT for delay time. Note 5: In column R.O.: G for general EMV, A for ambulance.

), some discussions are presented below. Firstly, few studies considered the integrated optimization of EMV-TPC strategies along the whole route simultaneously [71,84,105], which may be insufficient to meet the whole-process rescue demands of EMVs. Secondly, prioritizing EMVs will have a negative impact on associated SVs. Future studies should adopt suitable strategies to reduce the negative impact, such as adopting appropriate transition strategies, selecting smaller alternative rescue vehicles [106], and developing path-based or network-based algorithms based on selected routes. Thirdly, due to long transmission distances, the communication process between EMVs and TSCs or other SVs might be hindered by communication delays in the V2X environment [107]. However, limited studies have considered this factor [87,107,108].

6. Potential Directions for Future Research

This section highlights several research gaps that require attention based on our findings, as detailed below.

6.1. Uncovering Authentic Demand Characteristics through EMV Data Mining

Relying on overly strong model assumptions [28] and ignorance of the different characteristics of EMVs and SVs will significantly reduce the accuracy of models and algorithms, causing a disconnect between the optimized results and the real-world requirements. In the context of the big data era, future research should fully leverage the advantages of data and concentrate on uncovering the authentic rescue demand characteristics [26] and routing preferences [14,51,52] by collecting and mining actual EMV data (e.g., trajectory data, alarm data from the emergency department) to close the divide between the proposed algorithm and its real-world implementation [27,36].

6.2. Incorporating Distinct Characteristics of EMVs in EMV-RO Models

Apart from the single indicator of response times, future research in terms of EMV-RO methods can incorporate additional objectives [14,51,52], including minimizing driving risk, enhancing reliability, and mitigating negative impacts on SVs [71,72,90]. Concurrently, future research could select appropriate solving algorithms and incorporate EMVs’ distinct driving policies and vehicle attributes into the routing optimization model [99], which could augment the efficiency and accuracy of achieving optimal routes [98].

6.3. Implementing Active EMV-TPC Strategies

Compared with passive EMV-TPC strategies [76,82,83,85] (most are node-based strategies) activated by traditional detection means, future studies should endeavor to design active path-based or network-based EMV-TPC strategies [1,80,88] that prioritize EMVs along the rescue route while striving to minimize negative impacts on SVs. Moreover, the coupling problem between EMV-RO and EMV-TPC is also noteworthy as existing approaches have not fully addressed this matter.

6.4. Concentrating More on the Negative Impacts on SVs

Rather than exclusively optimizing EMVs, future studies should concentrate more on the negative impacts on SVs and select suitable indicators to quantify the negative impact, including SV waiting times, queue lengths, average number of stops, and passing capacity [71,72,86,87,90]. Moreover, future research can incorporate appropriate transition algorithms [93] and smaller rescue vehicles [106,112] in their models, which can help reduce the negative impacts and transition times [79,92]. The negative impact on pedestrians and non-motorized vehicles also needs to be considered [111].

6.5. Embracing Emerging Technologies in the Future Traffic Environment

The emergence of technologies such as UAV [49,50], Big Data [41,99], AI, and VANET [71] holds promise for substantially enhancing EMV rescue performance, particularly in terms of reducing response times and driving risk while minimizing the negative impacts on SVs. However, the performance of emerging technologies is also vulnerable to the challenges posed by adverse conditions, including equipment damage, communication delays, and adverse weather [76]. Additionally, most studies based on the IoV considered the ideal scenario where all vehicles are equipped with IoV and tested their proposed strategies in simulation software, which might not match real-world conditions and may not be practical to implement. Future research should utilize emerging technologies in EMV rescue while also considering the advantages and limitations of each technology for a comprehensive application. Moreover, empirical analysis should be utilized as much as possible in future research to verify the effectiveness of the proposed strategies in real-world scenarios.

7. Conclusions

This paper presented a comprehensive literature review aimed at investigating the theoretical methodologies and practical applications related to the optimization of routing and prioritizing for EMVs within the context of emerging technologies. We clarified the definition of the response time of EMVs according to the rescue process. With VOSviewer, we performed a bibliography analysis to investigate the present status of strategies for reducing response time in terms of the co-occurrence of keywords and co-citation of journals. According to the retrieved results, we categorized studies into three modules, including EMV-TTP, EMV-RO, and EMV-TPC. As with EMV-TTP, we divided the methods into model-based, simulation-based, API-based, and UAV-based methods. Regarding EMV-RO, we classified the algorithms into EMV-SRO, EMV-DRO, and EMV-ROAS algorithms. With regard to EMV-TPC, we divided the strategies into EMV-TPC strategies for intersections and road links. Finally, by identifying critical research gaps, this study recommended research suggestions to concentrate on uncovering authentic demand characteristics through EMV data mining, incorporating EMV distinct characteristics in EMV-RO models, implementing active EMV-TPC strategies, concentrating more on the negative impacts on SVs, and embracing emerging technologies.