Distributed Real-Time Vehicle Control for Emergency Vehicle Transit: A Scalable Cooperative Method

WenXi Wang; JunQi Zhang

cs.CVMar 26, 2026

Distributed Real-Time Vehicle Control for Emergency Vehicle Transit: A Scalable Cooperative Method

Authors: WenXi Wang, JunQi Zhang

Organizations: School of Computer Science and Technology, Key Laboratory of Embedded System and Service Computing, Ministry of Education, Tongji University, Shanghai 200092, China

Abstract

Rapid transit of emergency vehicles is critical for saving lives and reducing property loss but often relies on surrounding ordinary vehicles to cooperatively adjust their driving behaviors. It is important to ensure rapid transit of emergency vehicles while minimizing the impact on ordinary vehicles. Centralized mathematical solver and reinforcement learning are the state-of-the-art methods. The former obtains optimal solutions but is only practical for small-scale scenarios. The latter implicitly learns through extensive centralized training but the trained model exhibits limited scalability to different traffic conditions. Hence, existing methods suffer from two fundamental limitations: high computational cost and lack of scalability. To overcome above limitations, this work proposes a scalable distributed vehicle control method, where vehicles adjust their driving behaviors in a distributed manner online using only local instead of global information. We proved that the proposed distributed method using only local information is approximately equivalent to the one using global information, which enables vehicles to evaluate their candidate states and make approximately optimal decisions in real time without pre-training and with natural adaptability to varying traffic conditions. Then, a distributed conflict resolution mechanism is further proposed to guarantee vehicles' safety by avoiding their decision conflicts, which eliminates the single-point-of-failure risk of centralized methods and provides deterministic safety guarantees that learned methods cannot offer. Compared with existing methods, simulation experiments based on real-world traffic datasets demonstrate that the proposed method achieves faster decision-making, less impact on ordinary vehicles, and maintains much stronger scalability across different traffic densities and road configurations.

Figures & tables

Paper	Optimization Objective	Solution Method	Case Size	Algorithm Usage Condition	Control Paradigm	Training Time	Max Lane Count
Murray-Tuite et al.	EMV speed	Genetic algorithm	A segment with 20 OVs	None	Centralized	0	3
Hannoun et al.	EMV speed	CPLEX	A segment with 15 OVs	None	Centralized	0	3
Wu et al.	EMV speed, lane clearance cost	A* algorithm & commercial solvers	A segment with 54 OVs	None	Centralized	0	3
Lin et al.	EMV efficiency; state-change count	Customized algorithm & Gurobi	3 segments with 144 OVs	None	Centralized	0	3
Yang et al.	EMV efficiency; state-change count	Graph conv. soft actor-critic (GSAC)	A long segment with 80 OVs	Within 100 m ahead of EMV	Distributed	10–15 h	3
This paper	EMV efficiency; state-change count	Distributed real-time vehicle control	3 long segments with 436 OVs	None	Distributed	0	5

TABLE I: Comparison of Methods Based on Road Segment Vehicle Control Strategy

Fig. 1: Discretization of the considered roadway network.

Level	Speed (cells/s)	Speed (m/s)
0	0	0
1	1	6
2	2	12
3	3	18
4	4	24
5	5	30

TABLE II: SPEED LOOKUP TABLE FOR EMVs AND OVs

Panel of Figure 4 — Fig. 2: Illustration of prediction horizon $F_{j,n}$ .

Fig. 3: Illustration of a platoon. Vehicles in the platoon share the same lane and speed with consecutive positions. OV $n$ needs to check whether the platoon’s head ( $n_{k}$ ) or tail ( $n_{1}$ ) conflicts with other vehicles.

Fig. 4: Sensitivity analysis of weight parameters. (a) Mean total lane changes. (b) Mean versus standard deviation. The blue star indicates the optimal configuration.

Name	Type	Dimension	Domain
$T_{H}$	Integer variable	1	[3, 12]
$\triangle$	Integer variable	1	[1, $T_{H}$ )
$i^{t}_{n}$	Integer variable	$(T_{H},\|\mathcal{N}\|)$	[1, $I$ ]
$a^{t}_{n}$	Integer variable	$(T_{H}-1,\|\mathcal{N}\|)$	[0, 1]
$b^{t}_{n}$	Integer variable	$(T_{H}-1,\|\mathcal{N}\|)$	[0, 1]
$l^{t}_{n}$	Integer variable	$(T_{H},\|\mathcal{N}\|)$	[1, 3]

TABLE III: MAIN PARAMETERS USED IN COMPARISON ALGORITHMS

Parameter	Value
GSAC Training Environment
Number of lanes	3
Traffic density	110 veh/km
Average initial OV speed $\bar{V}_{\text{OV}}$	2
GSAC Network Architecture
FCN Encoder	2 layers, 32 units each

TABLE IV: PARAMETER SETTINGS OF GSAC_IDM

Case	The value of $f^{\prime}$					Total planning time ( $s$ )					Collision Rate (%)
Case	SDVC	GSAC_IDM	CRHA	CRHA_1	CRHA_2	SDVC	GSAC_IDM	CRHA	CRHA_1	CRHA_2	SDVC	GSAC_IDM	CRHA	CRHA_1	CRHA_2
30_14_5	16	27	13 /22.38/145	13 /16.17/59	13 /15.60/30	0.036	0.01	1.07/7.92/14.88	0.52/4.39/8.90	0.49/4.10/10.28	0	0	0	0	0
35_18_4	25	51	27/83.45/205	25 /31.00/88	25 /29.11/68	0.046	0.01	1.24/12.85/22.12	0.92/11.30/20.07	0.58/9.79/20.05	0	0	0	0	0
43_15_5	28	58	27 /96.89/253	28/52.81/133	28/36.59/49	0.036	0.01	2.04/12.71/18.13	1.38/11.52/20.05	0.82/10.56/18.09	0	0	0	0	0
47_18_4	25	61	28/100.67/292	32/54.91/175	31/51.70/152	0.043	0.01	2.46/13.91/20.92	2.15/12.03/16.49	1.19/11.71/20.08	0	6	0	0	0
54_24_3	14	24	14 /28.67/233	14 /19.82/55	14 /18.82/27	0.061	0.02	2.65/15.85/24.81	1.51/9.61/17.55	1.48/8.68/15.38	0	0	0	0	0

TABLE V: COMPARISON OF DIFFERENT ALGORITHMS

Fig. 5: Vehicle trajectories under the proposed SDVC in case 54_24_3.

Density (veh/km)	$\|N\|$	$\Delta v$	$f^{\prime}\downarrow$		Collision Rate (%) $\downarrow$		Decision Time (ms) $\downarrow$
Density (veh/km)	$\|N\|$	$\Delta v$	SDVC	GSAC_IDM	SDVC	GSAC_IDM	SDVC	GSAC_IDM
64	81	1	27	34	0	0	2.5	0.75
76	96	1	56	71	0	0	2.5	0.75
76	96	2	60	90	0	1	2.4	0.75
88	111	1	61	126	0	2	2.4	0.75
		2	66	178	0	4	2.5	0.75

TABLE VI: Performance Comparison Under Varying Traffic Density and Speed Heterogeneity

Fig. 6: Variation trends of

f^{\prime}

under different traffic densities and speed heterogeneity levels.

Lane Count	$\|N\|$	$f^{\prime}\downarrow$		Collision Rate (%) $\downarrow$		Decision Time (ms) $\downarrow$
Lane Count	$\|N\|$	SDVC	GSAC_IDM	SDVC	GSAC_IDM	SDVC	GSAC_IDM
3	141	125	245	0	6	2.7	0.75
4	192	156	304	0	7	2.7	0.75
5	248	150	288	0	4	2.7	0.75

TABLE VII: Performance Comparison Under Varying Lane Configurations

Intelligent Transportation Systems

Intelligent Transportation Systems apply sensing, communication, AI, and optimization to make transport networks safer and more efficient.

See all 297 papers

Mar 20, 2026

Emergency Preemption Without Online Exploration: A Decision Transformer Approach

Emergency vehicle (EV) response time is a critical determinant of survival outcomes, yet deployed signal preemption strategies remain reactive and uncontrollable. We propose a return-conditioned framework for emergency corridor optimization based on the Decision Transformer (DT). By casting corridor optimization as offline, return-conditioned sequence modeling, our approach (1) eliminates online environment interaction during policy learning, (2) enables dispatch-level urgency control through a single target-return scalar, and (3) extends to multi-agent settings via a Multi-Agent Decision Transformer (MADT) with graph attention for spatial coordination. On the LightSim simulator, DT reduces average EV travel time by 37.7% relative to fixed-timing preemption on a 4x4 grid (88.6 s vs. 142.3 s), achieving the lowest civilian delay (11.3 s/veh) and fewest EV stops (1.2) among all methods, including online RL baselines that require environment interaction. MADT further improves on larger grids, overtaking DT with 45.2% reduction on 8x8 via graph-attention coordination. Return conditioning produces a smooth dispatch interface: varying the target return from 100 to -400 trades EV travel time (72.4-138.2 s) against civilian delay (16.8-5.4 s/veh), requiring no retraining. A Constrained DT extension adds explicit civilian disruption budgets as a second control knob.

Haoran Su, Hanxiao Deng, Yandong Sun

New York University · University of California, Berkeley

Traffic Signal Control Decision Transformer Offline RL Multi-Agent Reinforcement Learning

Ask PDF

Dec 3, 2024

Highway Congestion Reduction through Reinforcement Learning Based Eulerian Headway Control

Connected automated vehicles (CAVs) equipped with adaptive cruise control (ACC) create new opportunities for highway congestion mitigation. Traditional practice relies on Eulerian variable speed limits (VSL) which regulate traffic through roadside signs, but suffer from infrequent updates and limited driver compliance. Recent research explored Lagrangian strategies that directly control individual vehicles, offering high reactivity and compliance, yet in realistic multi-lane settings they depend on drivers' latent lane-change intentions, making robust vehicle-level decisions difficult. Hence, we propose an Eulerian control system optimized through reinforcement learning, that (i) leverages ACC for reactivity and compliance, and (ii) obviates dependence on latent driver intentions by regulating aggregate density near bottlenecks, crucially via headway commands rather than speed commands. We evaluate two variants of our system, time-headway and distance-headway control, in large-scale simulations across a range of traffic conditions. Both variants outperform baselines, improving traffic flow by up to 10.6% over human traffic and 6.7% over traditional VSL. To strengthen evaluation, we propose a novel boundary-aware speed metric addressing a recognized flaw in simulation studies with dynamic vehicle entry and exit. The empirical results, together with our emphasis on deployable system design, suggest a path towards practical, safe, and scalable highway congestion mitigation.

Yaron Veksler, Sharon Hornstein, Han Wang +2

Research and Development Labs, General Motors, Herzliya, Israel. · Department of Civil and Environmental Engineering, UC Berkeley.

RL Control Intelligent Transportation Systems Autonomous Driving

Ask PDF

Jul 26, 2026

GNN-based Multi-Agent Control of Traffic Shockwaves in Sparse Vehicular Ad-hoc Networks

Traffic shockwaves are stop-and-go waves that propagate upstream through the streams of vehicles and are one of the major causes of traffic congestion, fuel inefficiency, and increased accident rates in modern transportation systems. Although Connected and Autonomous Vehicles (CAVs) offer a promising opportunity to mitigate such shockwaves, most existing control strategies rely on global traffic state information, making them impractical for early-stage deployment of Vehicular Ad-hoc Networks (VANETs). In this paper, we propose a decentralized Multi-Agent Reinforcement Learning (MARL) framework that integrates a Graph Neural Network (GNN) to enhance the control architecture of connected and autonomous vehicles. The proposed approach enables vehicles to learn cooperative control policies using locally available information and interaction with neighboring vehicles. The effectiveness of the proposed scheme is evaluated using a scalable simulation environment under realistic highway traffic conditions. Simulation results show that the proposed GNN-based MARL framework can reduce the propagation of traffic shockwaves by up to 80%, even when only 10% of the vehicles are connected.

Prachi Nandi, Madhuri Malakar, Sonakshi Satpathy +1

Microsoft · Dept. of Computer Science and Engineering GITAM University Bengaluru, Karnataka · NatWest Group +1

Multi-Agent Control Decentralized MARL

Ask PDF

Abstract

Intelligent Transportation Systems

Explore similar work

Emergency Preemption Without Online Exploration: A Decision Transformer Approach

Highway Congestion Reduction through Reinforcement Learning Based Eulerian Headway Control

GNN-based Multi-Agent Control of Traffic Shockwaves in Sparse Vehicular Ad-hoc Networks