Smart Traffic Optimization System for Emergency Vehicle Priority Based on SUMO and TraCI

Smart Trafc Optimization System for Emergency Vehicle Priority Based on SUMO and TraCI

1st Prof. Mayuri Shende

Department of Computer Science and Engineering Priyadarshini College of Engineering, Nagpur, India

2nd Akshata Titarmare

Department of Computer Science and Engineering Priyadarshini College of Engineering, Nagpur, India

3rd Rojel Sheikh

Department of Computer Science and Engineering Priyadarshini College of Engineering, Nagpur, India

4th Bhupesh Jambhulkar

Department of Computer Science and Engineering Priyadarshini College of Engineering, Nagpur, India

5th Radhey Talware

Department of Computer Science and Engineering Priyadarshini College of Engineering, Nagpur, India

6th Shivam Dhule

Department of Computer Science and Engineering Priyadarshini College of Engineering, Nagpur, India

Abstract – Emergency vehicles (EVs) such as ambulances, re trucks, and police vehicles frequently experience delays at signal- ized intersections because conventional xed-time trafc signal controllers do not account for their presence. These delays can be life-threatening in time-critical situations. This paper proposes a lightweight, simulation-rst framework for emergency vehicle priority using the Simulation of Urban Mobility (SUMO) platform coupled with the Trafc Control Interface (TraCI). The system detects approaching emergency vehicles in real time, computes a priority score based on distance, estimated arrival time, and downstream queue conditions, and dynamically pre- empts the affected trafc signal. A green-wave coordination mechanism extends priority across successive intersections along the emergency route, while a queue-aware and spillback-aware recovery strategy limits disruption to regular trafc. The frame- work is evaluated conceptually across four trafc-density scenar- ios using SUMO-based experimental settings. Illustrative results are presented to demonstrate expected trends in emergency vehicle travel time, waiting time, and normal trafc delay. The proposed approach is intended as a low-cost, infrastructure-light alternative to expensive V2X-dependent systems for smart-city trafc management.

Index TermsSUMO, TraCI, Emergency Vehicles, Trafc Signal Control, Green Wave, Intelligent Transportation Systems, Queue Management

1. Introduction

   Urban trafc congestion has become one of the most per- sistent challenges faced by modern cities. Rapid urbanization, increasing vehicle ownership, and limited road infrastructure expansion have resulted in trafc networks that regularly operate near or above capacity during peak hours. While congestion is inconvenient for ordinary commuters, it becomes

   critical when it delays emergency vehicles (EVs) such as ambulances, re trucks, and police vehicles. Every additional minute of delay for an ambulance responding to a cardiac arrest or a re truck responding to a structural re can materi- ally affect patient outcomes and property damage. Emergency medical service literature has long established that response time is one of the strongest predictors of survival in time- critical incidents, which makes the reduction of EV travel delay a matter of direct public safety signicance rather than mere trafc efciency.

   The majority of urban intersections worldwide are still controlled by xed-time or simple actuated signal plans. Fixed- time control allocates green time according to pre-computed cycle plans that do not respond to real-time trafc conditions, let alone to the presence of an approaching emergency vehicle. Even actuated controllers, which adjust green duration based on loop-detector occupancy, generally lack any mechanism to recognize an EV as a distinct, high-priority class of trafc. As a result, EVs are frequently forced to queue behind regular vehicles, execute risky maneuvers such as crossing against a red signal, or rely on siren-based informal right-of-way negotiation with surrounding drivers, all of which introduce safety risk and unpredictable delay.

   To address this limitation, trafc engineering has historically developed emergency vehicle pre-emption systems, in which an approaching EV can override the normal signal sequence and force a green phase in its direction of travel. Early pre- emption technologies relied on line-of-sight optical or acoustic emitters mounted on the vehicle and receivers mounted at

   the intersection. While effective at a single intersection, such systems are inherently local: they do not coordinate priority across multiple consecutive intersections, and they do not account for the state of trafc queues downstream of the EVs path. This motivates the concept of green-wave trafc management, in which a sequence of signals along an arterial is coordinated so that a vehicle traveling at a target speed encounters a continuous progression of green phases. Green- wave concepts have traditionally been applied to regular trafc ow optimization, but their extension to a single, dynamically routed emergency vehicle requires additional real-time co- ordination logic that classical ofine bandwidth-optimization methods do not provide.

   The emergence of Intelligent Transportation Systems (ITS) has opened new possibilities for solving this problem com- putationally rather than purely through roadside hardware. Modern ITS research increasingly relies on microscopic trafc simulation to design, test, and validate control algorithms before any real-world deployment is attempted. Among the available simulation platforms, the open-source Simulation of Urban Mobility (SUMO) tool has become a widely used research standard because it supports large-scale, continuous- space, continuous-time microscopic simulation of individual vehicles, including explicit support for emergency vehicle classes. SUMO exposes a run-time control mechanism called the Trafc Control Interface (TraCI), which allows an external control program to query and modify the state of a running simulation at every simulation step. TraCI therefore makes it possible to implement and test EV-detection, signal pre- emption, and green-wave coordination logic entirely in soft- ware, without requiring physical roadside sensors or vehicle- to-infrastructure (V2I) hardware during the design and valida- tion phase.

   Despite the growing body of research in this area, several gaps remain. Many existing pre-emption systems assume that only a single emergency vehicle is present in the network at any given time, which does not reect the reality of large cities where multiple EVs may be active concurrently. A num- ber of proposed solutions depend on expensive infrastructure such as roadside units, dedicated short-range communication (DSRC), or full V2X connectivity, which limits their near- term deployability, particularly in resource-constrained cities. Reinforcement-learning-based trafc signal controllers have shown strong performance in general trafc optimization, but many such models are computationally heavy, require extensive training data, and are not specically designed around the coupling of EV routing and signal pre-emption. Finally, relatively few studies explicitly model the effect of pre-emption on downstream queue formation and spillback, which can create secondary congestion that ultimately slows the EV itself or worsens conditions for regular trafc once the EV has passed.

   The motivation of this work is therefore to design a simulation-rst framework that is lightweight enough to be evaluated without specialized infrastructure, yet technically complete enough to address EV detection, dynamic signa

   pre-emption, multi-intersection green-wave coordination, and queue/spillback-aware recovery within a single coherent sys- tem. The research problem addressed in this paper can be summarized as follows: how can a trafc control framework built on SUMO and TraCI simultaneously reduce emergency vehicle delay, extend priority coordination across successive intersections, and limit the resulting disruption to normal trafc, without relying on expensive external communication infrastructure? The research objectives that follow from this problem are to (i) model realistic emergency vehicle behavior within a microscopic simulation environment, (ii) design a TraCI-based detection and priority-scoring mechanism, (iii) implement dynamic signal pre-emption combined with green- wave coordination, and (iv) incorporate queue-aware and spillback-aware safeguards for normal trafc.

   The main contributions of this paper are:

   1. A lightweight, simulation-rst SUMOTraCI framework that unies emergency vehicle detection, priority scor- ing, and dynamic signal pre-emption within a single control loop.

   2. A distance-, arrival-time-, and queue-aware priority scor- ing formulation suitable for multi-vehicle emergency scenarios.

   3. A green-wave coordination mechanism that extends sig- nal priority across successive intersections along the projected route of an emergency vehicle.

   4. A queue-aware and spillback-aware recovery strategy that limits the disruption caused to regular trafc after pre-emption ends.

2. Literature Review

   This section reviews prior work across six themes relevant to the proposed system: emergency vehicle signal pre-emption, emergency vehicle routing, adaptive trafc signal con- trol, IoV/V2X-based emergency management, reinforcement- learning-based trafc control, and queue-aware/spillback- aware control.

   1. Emergency Vehicle Signal Pre-emption

      Foundational pre-emption research modeled the transition between normal and pre-emptive signal states and proposed control strategies to reduce disruption to ordinary trafc during and after pre-emption [13]. More recent work has coupled pre- emption logic with software-dened and centralized trafc- management architectures that manage the full emergency response lifecycle, from incident detection to signal recovery [10]. Microscopic behavior of EVs, including special driving rights and interaction with surrounding trafc, has also been explicitly modeled within SUMO [8], and validated through trajectory-level analysis of EVtrafc interaction [9].

   2. Emergency Vehicle Routing

      Dynamic routing research has largely built on shortest-path formulations, beginning with classical dispatching and routing models for disaster and emergency response [24] and dynamic ambulance routing models supporting multiple simultaneous

      responses [25]. Metaheuristic approaches such as NSGA-II and MOPSO have been applied to disaster-response ambulance routing with variable patient condition [26], while simulation- based routing models incorporating patient medical severity have also been proposed [27]. Safety-aware dynamic shortest- path computation has further been studied using data-driven optimization [28].

   3. Adaptive Trafc Signal Control

      Long before learning-based methods, adaptive systems such as SCOOT [18] and SCATS [19] demonstrated that real- time detector data could be used to adjust cycle length, split, and offset in response to measured trafc conditions. The demand-responsive OPAC strategy extended this con- cept with an optimization-based phase-selection approach [20]. Arterial-level coordination has been studied through bandwidth-maximization models such as MAXBAND [17] and its asymmetrical multi-band extension [15], as well as recent Euclidean-inspired green-wave theory for arbitrary ar- terial networks [16].

   4. IoV/V2X-Based Emergency Management

      The Internet of Vehicles (IoV) paradigm, built on VANET and IoT convergence, has been surveyed as a foundation for vehicle-to-infrastructure communication in intelligent trafc management [30], and applied specically to city-wide intel- ligent trafc management [29]. Such approaches offer strong theoretical performance but generally assume the availability of dedicated communication infrastructure across the network.

   5. Reinforcement-Learning-Based Trafc Control

      Deep reinforcement learning (DRL) has been widely applied to trafc signal control. Early work used a discrete trafc- state encoding with convolutional Q-learning within SUMO [22], later extended with asynchronous n-step Q-learning [31]. Policy-gradient and value-function-based RL methods [32], graph-convolutional-network-based state representations [33], and experience-replay-based DQN methods [35] have all been explored. Coordinated multi-agent RL approaches

      [36] and intersection-level intelligent control frameworks such as IntelliLight [23] have further improved network-level per- formance. EMVLight specically couples EV routing with decentralized RL-based signal pre-emption [11], [12].

   6. Queue-Aware and Spillback-Aware Control

   Queue spillback has been identied as a major cause of arterial gridlock, motivating connected-vehicle-based spillback detection and control [14]. High-resolution, event-based de- tector data has been proposed as a means of improving both modeling and control accuracy in congested conditions [34]. These queue- and spillback-aware principles are directly relevant to limiting the disruption that EV pre-emption can cause to regular trafc.

3. Literature Review Table

   Table I summarizes 38 representative studies spanning the six themes discussed above.

   Synthesis and research gap: The reviewed literature shows steady progress across four largely separate threads: simulation platforms and vehicle dynamics [1][7], [37], [38]; EV-specic

   pre-emption and routing [8][10], [13], [24][28]; general adaptive and learning-based signal control [15][23], [31] [33], [35], [36]; and infrastructure-dependent IoV/V2X or

   connected-vehicle methods [14], [29], [30]. Very few works combine EV detection, multi-intersection green-wave coordi- nation, and explicit queue/spillback safeguards within a single lightweight, simulation-rst framework that does not assume V2X connectivity or heavy RL training. This gap motivates the framework proposed in the following sections.

4. Research Gap and Proposed Contribution

   Existing approaches to emergency vehicle priority com- monly exhibit one or more of the following limitations: (i) single-EV assumptions that break down when multiple emer- gency vehicles operate concurrently; (ii) limited coordination across more than one intersection, restricting benet to a purely local pre-emption event; (iii) dependence on expensive infrastructure such as dedicated detectors, roadside units, or full V2X connectivity, which slows real-world adoption; (iv) high computational requirements associated with training and deploying reinforcement-learning controllers; and (v) weak or absent handling of queue length and spillback effects, allowing pre-emption to create secondary congestion for regular trafc. To address these limitations, this paper proposes a lightweight SUMO + TraCI simulation-rst framework that combines emergency vehicle detection, adaptive signal pre- emption, green-wave coordination, and queue-aware trafc management within a single control loop. The framework is deliberately designed to avoid dependence on external communication infrastructure during the design and evaluation stage: all state information (vehicle position, speed, edge, and queue length) is obtained directly from the SUMO simulatio through TraCI, which mirrors the type of information that would otherwise be obtained from roadside sensors or GPS- based tracking in a real deployment. This design keeps the ap- proach computationally light relative to RL-based controllers, while still supporting multiple concurrent emergency vehicles through a priority-scoring mechanism, and while explicitly limiting the impact of pre-emption on downstream congestion

   through queue- and spillback-aware recovery logic.

5. Methodology

   1. SUMO Network Conguration

      The proposed framework is evaluated on a SUMO road network consisting of multiple signalized intersections con- nected by arterial and side-street links. The network is dened using standard SUMO input les: a network le (.net.xml) describing edges, lanes, and trafc-light logic; a route le (.rou.xml) describing vehicle types and demand; and an additional le describing detectors where required. Trafc

      TABLE I: Summary of Reviewed Literature

      No. Author/Year Approach Key Result Limitation

      1. Behrisch et al., 2011 [1] SUMO platform overview Established open-source microscopic simulation

         base

         No EV-specic priority logic

      2. Krajzewicz et al., 2012 [2] SUMO applications overview Documented growing SUMO application areas Descriptive, not a control method

      3. Lopez et al., 2018 [3] SUMO/TraCI microscopic simulation Widely adopted simulation-control coupling General-purpose, not EV-specic

      4. Wegener et al., 2008 [4] TraCI interface design Enabled online control of running simulations Interface only, no control policy

      5. Krauß, 1998 [5] Collision-free car-following model Basis for SUMOs default vehicle dynamics No signal or priority modeling

      6. Gipps, 1981 [6] Behavioural car-following model Inuential safe-gap driving model Predates microscopic ITS tools

      7. Treiber et al., 2000 [7] Intelligent Driver Model (IDM) Realistic congested-ow reproduction Not EV-aware

      8. Bieker-Walz et al., 2018 [8] EV modeling in SUMO Special EV rights and rescue-lane formation Limited to single-EV scenario

      9. Corte´s et al., 2023 [9] EMV trajectorytrafc interaction Realistic EMV/POV interaction modeling Focused on trajectory realism, not

         signals

      10. Bagheri & Ferrari, 2023 [10] Software-dened preemption Faster EMS response via centralized control Requires OMNeT++/VEINS net-

          work stack

          demand is generated as a mixture of background (regular) trafc and a smaller number of emergency vehicle trips injected at controlled intervals. Signal conguration follows a standard multi-phase scheme with protected and permitted movements, and a base cycle length representative of a typical urban intersection.

          Fig. 1: Proposed system architecture.

   2. Emergency Vehicle Modelling

      Emergency vehicles (ambulance, re truck, police vehicle) are modeled as a distinct SUMO vehicle class with elevated maximum speed and acceleration relative to regular trafc, and are assigned the emergency vehicle class so that SUMOs built-in special-vehicle behavior (e.g., other vehicles yielding) is available. Each EV is associated with a pre-dened route and a detection zone, dened as the set of edges within a congurable distance upstream of each intersection along that route. A minimal example of an EV type denition is shown below.

      <vType id=”emergency” vClass=”emergency” guiShape=”emergency” length=”6.5″ maxSpeed=”25″ accel=”3.0″ decel=”6.0″ color=”1,0,0″/>

      <vehicle id=”ev_1″ type=”emergency” depart=”0″ route=”ev_route_1″/>

   3. Emergency Vehicle Detection

      At every simulation step, the TraCI client queries the set of active vehicles and lters those whose vClass equals emergency. For each detected EV, the controller retrieves

      its identier, current position, speed, current edge, the next signalized intersection along its route, and the remaining distance to that intersection using TraCIs vehicle- and route- related calls. This information is refreshed at every control step so that the priority mechanism always operates on the most recent EV state.

   4. Priority Decision Mechanism

      When more than one EV or competing demand is present, the controller computes a priority score P for each approach- ing EV using

      P = w1D1 + w2T 1 + w3Q + w4E (1)

      where D is the distance from the EV to the intersection, T is the estimated time of arrival at the intersection, Q is the current queue length on the EVs approach (in vehicles), E is a binary emergency-status indicator (E = 1 for a conrmed active emergency, E = 0 otherwise), and w1,…, w4 are non-negative weights satisfying w1 + w2 + w3 + w4 = 1, tuned so that closer and sooner-arriving EVs receive higher priority while heavier downstream queues are also taken into account. The intersection with the highest aggregate priority score among competing requests is served rst.

   5. Dynamic Signal Pre-emption

      When an EV enters the detection zone of an intersection, the controller checks whether the current signal phase already serves the EVs direction of travel. If not, the controller inserts a transition phase (an intergreen/clearance interval consistent with the existing phase plan) followed by a green phase for the EVs approach, subject to two safety constraints:

      (i) a minimum clearance time must elapse before switching away from any phase currently displaying green to conicting movements, and (ii) the minimum green time of the phase being interrupted must have already elapsed. These constraints prevent unsafe abrupt phase changes.

   6. Green-Wave Coordination

      Once an EVs route and expected arrival time at the next intersection are known, the controller estimates its expected ar- rival time at each subsequent intersection along the route using the EVs nominal speed and the distance between intersections. Pre-emption requests are then scheduled in advance at each

      Algorithm 1 TraCI-Based EV Priority Control Loop

      1: Connect to SUMO via TraCI

      2: while simulation has active vehicles do

      3: Advance simulation by one step

      4: EV s set of active vehicles with class emergency

      5: for each ev EV s do

      6: if ev is within detection zone of next intersection

      then

      7: Compute P for ev using Eq. (1)

      8: end if

      9: end for

      10: Rank pending pre-emption requests by P

      11: for each intersection with a pending request do

      12: if safety constraints satised then

      13: Apply/extend green phase for highest-priority EV

      14: Schedule downstream green-wave requests

      15: end if

      16: end for

      17: for each pre-empted intersection do

      18: if queue on non-priority approach exceeds threshold

      then

      19: Shorten recovery interval for that approach

      20: end if

      21: end for

      22: end while

      downstream intersection so that, ideally, the EV encounters a continuous sequence of green phases without stopping, similar in spirit to conventional green-wave progression but computed dynamically for the EVs specic route rather than for a xed arterial timing plan.

   7. Queue and Spillback Management

      To limit the impact of pre-emption on regular trafc, the controller monitors queue length on all approaches at each pre-empted intersection using TraCI-reported halting-vehicle counts. If a non-priority approach queue exceeds a cong- urable spillback threshold (i.e., approaches the physical storage capacity of the link), the controller shortens the subsequent pre-emption-recovery interval for that approach so that it receives green time sooner, reducing the likelihood of the queue extending into the upstream intersection.

   8. TraCI Control Algorithm

   Algorithm 1 summarizes the core per-step control loop.

6. Experimental Setup

   The proposed framework is intended to be evaluated in SUMO under four trafc-density scenarios, summarized in Table II. These values represent simulation settings and ex- perimental assumptions rather than measured eld data.

   Each simulation run is dened by the following assumed parameters: simulation duration of 3600 s (one hour) per run; a network of 46 signalized intersections along a single arterial corridor; 24 emergency vehicle trips injected at randomized

   TABLE II: Simulation Trafc-Density Scenarios

   Scenario Trafc Density

   Low 100 vehicles/hour

   Medium 300 vehicles/hour

   High 600 vehicles/hour Very High 900 vehicles/hour

   intervals per run; background trafc generated according to the density levels in Table II; a base signal cycle length of 90 s with a 4-phase plan per intersection; an EV detection distance of 300 m upstream of each intersection; and an EV free- ow speed of 6080 km/h subject to the safety-constrained car-following model. To account for stochastic variability in vehicle generation and driver behavior, it is recommended that each scenario be repeated using at least 10 independent random seeds, with results reported as mean values together with standard deviation.

7. Results and Discussion

   Because this paper does not include eld-collected or pre- viously executed SUMO experimental logs, Table III presents Illustrative Simulation Results: a table structure with rep- resentative, clearly labelled placeholder values intended to show the expected trend across methods and densities. These values are illustrative only and must be replaced with actual SUMO/TraCI measurement outputs prior to any claim of measured performance.

   Fig. 2: Working on Sumo Environment

   TABLE III: Illustrative Simulation Results (values are illus- trative and must be replaced with actual SUMO results)

   Density	Method	EV Travel (s)	EV (s)	Wait	Normal Delay (s)
   Low	Fixed-Time
   Low	Proposed
   Medium	Fixed-Time
   Medium	Proposed
   High	Fixed-Time
   High	Proposed
   Very High	Fixed-Time
   Very High	Proposed

   Based on the design of the proposed mechanism, several qualitative trends are expected once actual SUMO exper-

   iments are run. EV travel and waiting time are expected to decrease relative to xed-time and conventional single- intersection pre-emption, with the largest relative benet an- ticipated at medium-to-high trafc densities where queuing is signicant enough for coordination to matter, but not so severe that downstream links are already saturated. As trafc density increases toward the very high scenario, the benet of green-wave coordination is expected to diminish because downstream links may already be near capacity regardless of signal timing, and queue-aware recovery becomes increasingly important to prevent spillback. The effect on normal (non- priority) trafc delay is expected to be small but non-zero immediately after an EV passes, with the queue-aware recov- ery mechanism intended to shorten this transient compared with conventional pre-emption. Signal recovery the time for a pre-empted intersection to resume its normal phase plan is expected to be faster under the proposed method due to the explicit recovery-interval adjustment described in Section V-G. Scalability to a larger number of concurrent EVs is expected to depend primarily on the priority-scoring mechanism correctly sequencing competing requests rather than serving them in arrival order alone. All of these expectations should be treated as hypotheses to be conrmed through actual SUMO/TraCI experimentation rather than established ndings.

   Fig. 3: Maps in Project

8. Applications and Limitations

   The proposed framework is applicable to ambulance dis- patch, re and rescue response, police pursuit and es- cort operations, disaster-response trafc management, and broader smart-city Intelligent Transportation System deploy- ments where reducing emergency response time is a policy pri- ority. At the same time, several limitations should be acknowl- edged. First, there is an inherent simulation-to-reality gap: SUMOs car-following and lane-changing models approximate but do not perfectly reproduce real driver behavior. Second, the systems effectiveness depends on accurate, low-latency vehicle detection, which in a real deployment would require reliable GPS or roadside sensing rather than the idealized state access provided by TraCI in simulation. Third, the trafc- demand values used in Section VI are assumptions rather than calibrated eld data, and results will vary with real demand patterns. Fourth, real-world communication delays between vehicles, controllers, and any backend system are not modeled here and could reduce the practical benet of pre-emption.

   Fifth, driver behavior in the presence of an approaching EV (e.g., informal yielding, rescue-lane formation) is simplied relative to real, heterogeneous driver responses. Finally, real- world testing and pilot deployment would be required before the framework could be considered validated for operational use.

9. Future Scope

   Future extensions of this work include incorporating Vehicle-to-Everything (V2X) communication and 5G/6G con- nectivity for lower-latency EV detection; using IoT roadside sensors and GPS-based ambulance tracking to replace ideal- ized simulation state access; exploring CCTV/image-based EV detection at intersections; integrating machine learning and reinforcement learning to adaptively tune the priority-score weights in Eq. (1); building a digital-twin representation of a real city network for higher-delity validation; extending the framework to explicitly coordinate multiple simultaneous EVs across overlapping routes; and interfacing the control logic with real-world trafc-signal controller hardware for eld trials.

10. Conclusion

This paper presented a lightweight, simulation-rst frame- work for emergency vehicle priority based on SUMO and TraCI. Motivated by the persistent delay experienced by ambulances, re trucks, and police vehicles at signal- controlled intersections under xed-time and conventional pre-emption schemes, the proposed approach combines real- time emergency vehicle detection, a distance-, arrival-time-

, and queue-aware priority scoring mechanism, dynamic signal pre-emption with explicit safety constraints, green- wave coordination across successive intersections, and queue- aware, spillback-aware recovery to limit disruption to regular trafc. A compact literature review of 38 studies across eergency vehicle pre-emption, emergency vehicle rout- ing, adaptive trafc signal control, IoV/V2X-based emer- gency management, reinforcement-learning-based control, and queue/spillback management was used to identify a gap: the absence of a unied, infrastructure-light, simulation-rst framework addressing multi-EV coordination together with downstream trafc protection. A methodology addressing this gap was described in detail, together with an experimental de- sign spanning four trafc-density scenarios and three baseline comparisons. Because this paper does not include executed SUMO experiments, results were presented explicitly as illus- trative placeholders rather than measured ndings, consistent with responsible reporting practice. Expected benets of the proposed approach include reduced emergency vehicle travel and waiting time, faster post-pre-emption signal recovery, and bounded impact on normal trafc, particularly under medium- to-high congestion. Future work will focus on executing the described SUMO/TraCI experiments, calibrating the priority- score weights, and progressing toward real-world pilot deploy- ment in coordination with local trafc authorities.

References

1. M. Behrisch, L. Bieker, J. Erdmann, and D. Krajzewicz, SUMO Simulation of Urban MObility: An Overview, in Proc. SIMUL 2011, 3rd Int. Conf. Adv. Syst. Simul., Barcelona, Spain, 2011, pp. 6368.

2. D. Krajzewicz, J. Erdmann, M. Behrisch, and L. Bieker, Recent Devel- opment and Applications of SUMO Simulation of Urban MObility, Int. J. Adv. Syst. Meas., vol. 5, no. 3&4, pp. 128138, 2012.

3. P. A. Lopez, M. Behrisch, L. Bieker-Walz, J. Erdmann, Y.-P. Flo¨ttero¨d,

   R. Hilbrich, L. Lu¨cken, J. Rummel, P. Wagner, and E. Wießner, Microscopic Trafc Simulation using SUMO, in Proc. IEEE 21st Int. Conf. Intell. Transp. Syst. (ITSC), Maui, HI, USA, 2018, pp. 25752582.

4. A. Wegener, M. Pio´rkowski, M. Raya, H. Hellbru¨ck, S. Fischer, and J.-P. Hubaux, TraCI: An Interface for Coupling Road Trafc and Network Simulators, in Proc. 11th Commun. Netw. Simul. Symp. (CNS), New York, NY, USA, 2008, pp. 155163.

5. S. Krauß, Microscopic Modeling of Trafc Flow: Investigation of Col- lision Free Vehicle Dynamics, Ph.D. dissertation, German Aerospace Center (DLR), 1998.

6. P. G. Gipps, A behavioural car-following model for computer simula- tion, Transp. Res. Part B, vol. 15, no. 2, pp. 105111, 1981.

7. M. Treiber, A. Hennecke, and D. Helbing, Congested trafc states in empirical observations and microscopic simulations, Phys. Rev. E, vol. 62, no. 2, pp. 18051824, 2000.

8. L. Bieker-Walz et al., Analysis of the trafc behavior of emergency vehicles in a microscopic trafc simulation, presented at the SUMO User Conf., 2018.

9. A. Corte´s et al., Trajectory Simulation of Emergency Vehicles and Interactions with Surrounding Trafc, J. Adv. Transp., vol. 2023, Art. no. 5995950, 2023.

10. N. Bagheri and G. Ferrari, Software-dened trafc light preemption for faster emergency medical service response in smart cities, Eur. J. Oper. Res., vol. 310, no. 2, pp. 712736, 2023.

11. H. Su, Y. D. Zhong, B. Dey, and A. Chakraborty, EMVLight: A Decentralized Reinforcement Learning Framework for Efcient Passage of Emergency Vehicles, in Proc. AAAI Conf. Artif. Intell., vol. 36, 2022,

    pp. 45934601.

12. H. Su, Y. D. Zhong, J. Y. J. Chow, B. Dey, and L. Jin, EMVLight: A multi-agent reinforcement learning framework for an emergency vehicle decentralized routing and trafc signal control system, Transp. Res. Part C, vol. 146, Art. no. 103955, 2023.

13. X. Qin and A. M. Khan, Control strategies of trafc signal timing transition for emergency vehicle preemption, Transp. Res. Part C, vol. 25, pp. 117, 2012.

14. E. Christofa, J. Argote, and A. Skabardonis, Arterial Queue Spillback Detection and Signal Control Based on Connected Vehicle Technology, Transp. Res. Rec., vol. 2356, no. 1, pp. 6876, 2013.

15. C. Zhang, Y. Xu, N. H. Gartner, C. Stamatiadis, and T. Arsava, AM- Band: An Asymmetrical Multi-Band Model for Arterial Trafc Signal Coordination, Transp. Res. Part C, vol. 58, pp. 515531, 2015.

16. M. H. Friedman, B. L. Mark, and N. H. Gartner, Euclidean Approach to Green-Wave Theory Applied to Trafc Signal Networks, J. Adv. Transp., 2025.

17. J. D. C. Little, M. D. Kelson, and N. H. Gartner, MAXBAND: A program for setting signals on arteries and triangular networks, Transp. Res. Rec., no. 795, 1981.

18. P. B. Hunt and D. I. Robertson, The SCOOT on-line trafc signal optimisation technique, Trafc Eng. Control, vol. 23, no. 4, pp. 190 192, 1982.

19. A. G. Sims and K. W. Dobinson, The Sydney Coordinated Adaptive Trafc (SCAT) System Philosophy and Benets, IEEE Trans. Veh. Technol., vol. 29, no. 2, pp. 130137, 1980.

20. N. H. Gartner, OPAC: A Demand-Responsive Strategy for Trafc Signal Control, Transp. Res. Rec., no. 906, pp. 7581, 1983.

21. K. A. Diaz, D. Dailisan, U. Sharaf, C. Santos, Q. Gan, F. A. Uy,

    M. T. Lim, and A. M. Bayen, Adaptive Coordination Offsets for Signalized Arterial Intersections using Deep Reinforcement Learning, arXiv:2008.02691, 2020.

22. W. Genders and S. Razavi, Using a Deep Reinforcement Learning Agent for Trafc Signal Control, arXiv:1611.01142, 2016.

23. H. Wei, G. Zheng, H. Yao, and Z. Li, IntelliLight: A Reinforcement Learning Approach for Intelligent Trafc Light Control, in Proc. 24th ACM SIGKDD Int. Conf. Knowl. Discov. Data Mining, 2018, pp. 2496 2505.

24. A. Jotshi, Q. Gong, and R. Batta, Dispatching and routing of emergency vehicles in disaster mitigation using data fusion, Socio-Econ. Plan. Sci., vol. 43, no. 1, pp. 124, 2009.

25. S. Yoon and L. A. Albert, A dynamic ambulance routing model with multiple response, Transp. Res. Part E, vol. 133, Art. no. 101807, 2020.

26. M. Rabbani, N. Oladzad-Abbasabady, and N. Akbarian-Saravi, Ambu- lance routing in disaster response considering variable patient condition: NSGA-II and MOPSO algorithms, J. Ind. Manag. Optim., vol. 18, no. 2, pp. 10351062, 2022.

27. M. Sutherland and R. K. Chakrabortty, An optimal ambulance routing model using simulation based on patient medical severity, Healthc. Anal., vol. 1, no. 4, Art. no. 100256, 2023.

28. S. Jiang, Y. Zhang, R. Liu, M. Jafari, and M. Kharbeche, Data- driven optimization for dynamic shortest path problem considering trafc safety, IEEE Trans. Intell. Transp. Syst., vol. 23, no. 10, pp. 1823718252, 2022.

29. S. A. Elsagheer Mohamed and K. A. AlShalfan, Intelligent Trafc Management System Based on the Internet of Vehicles (IoV), J. Adv. Transp., vol. 2021, 2021.

30. J. Contreras-Castillo, S. Zeadally, and J. A. Guerrero-Iba´nez, Internet of Vehicles: Architecture, Protocols, and Security, IEEE Internet Things J., vol. 5, no. 5, pp. 37013709, 2018.

31. W. Genders and S. Razavi, Asynchronous n-step Q-learning adaptive trafc signal control, J. Intell. Transp. Syst., vol. 23, no. 4, pp. 319331, 2019.

32. S. S. Mousavi, M. Schukat, and E. Howley, Trafc light control using deep policy-gradient and value-fuction-based reinforcement learning, IET Intell. Transp. Syst., vol. 11, no. 7, pp. 417423, 2017.

33. T. Nishi, K. Otaki, K. Hayakawa, and T. Yoshimura, Trafc Signal Control Based on Reinforcement Learning with Graph Convolutional Neural Nets, in Proc. IEEE Int. Conf. Intell. Transp. Syst. (ITSC), 2018,

    pp. 877883.

34. X. Wu and H. X. Liu, Using high-resolution event-based data for trafc modeling and control: An overview, Transp. Res. Part C, vol. 42, pp. 2843, 2014.

35. J. Gao, Y. Shen, J. Liu, M. Ito, and N. Shiratori, Adaptive trafc signal control: Deep reinforcement learning algorithm with experience replay and target network, arXiv:1705.02755, 2017.

36. E. van der Pol and F. A. Oliehoek, Coordinated Deep Reinforcement Learners for Trafc Light Control, presented at the NIPS Workshop Learn. Multiagent Syst., 2016.

37. M. Behrisch, D. Krajzewicz, P. Wagner, and Y.-P. Wang, Comparison of Methods for Increasing the Performance of a DUA Computation, in Proc. DTA2008 Int. Symp. Dyn. Trafc Assignment, Leuven, Belgium, 2008.

38. D. Krajzewicz, E. Brockfeld, J. Mikat, J. Ringel, C. Ro¨ssel, W. Tuch- scheerer, P. Wagner, and R. Woesler, Simulation of Modern Trafc Lights Control Systems using the Open Source Trafc Simulation SUMO, in Proc. 3rd Ind. Simul. Conf., Berlin, Germany, 2005, pp. 299302.