Enhancing the Performance of Emergency Vehicles Through Reinforcement Learning Based DVSL Control Systems

Bandarian, Fatemeh; Gyamfi, Eric; Mitsakis, Evangelos; Mintsis, Evangelos; Malekjafarian, Abdollah; Golpayegani, Fatemeh · 2025 · Crossref

DOI: 10.1007/978-3-032-04774-8_130

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the challenge of optimizing traffic flow to prioritize emergency vehicles (EVs) in congested highway environments, specifically at weaving sections involving ramps. While Variable Speed Limits (VSL) and Deep Reinforcement Learning (DRL) have been successfully applied to general traffic management, existing models typically fail to account for the specific needs of EVs, which require rapid passage to reduce fatality rates. The authors propose a Differential Variable Speed Limit (DVSL) control system that dynamically adjusts speed limits across different lanes to facilitate EV movement while maintaining overall traffic efficiency. The study employs a Deep Deterministic Policy Gradient (DDPG) algorithm, an actor-critic reinforcement learning model, to train the DVSL controller. The system operates within a Markov Decision Process framework where the state is defined by occupancy rates from loop detectors, and the action consists of setting distinct speed limits for each lane. A novel reward function is designed to minimize the accumulated time loss of EVs, defined as the time lost due to driving slower than desired. The experimental setup utilizes the Simulation of Urban MObility (SUMO) platform to simulate a three-lane freeway section with an on-ramp and off-ramp. The traffic mix includes heavy vehicles (20%), passenger cars (79%), and emergency vehicles (1%). The simulation assumes a high penetration of Connected Automated Vehicles (CAVs) at 99%, which comply with dynamic speed limits, while EVs operate independently of these limits. The agents were trained over 20 episodes of two-hour simulations. The results demonstrate that the proposed DDPG-based DVSL system significantly outperforms a baseline scenario with no VSL control (fixed 65 mi/h speed limit). The most notable finding is a 32.4% reduction in the average accumulated time loss for emergency vehicles. Additionally, the system improved other key metrics, including a decrease in the average accumulated waiting time for EVs and a reduction in the average travel time for all vehicle types. These improvements indicate that the dynamic control of speed limits effectively clears paths for emergency vehicles without severely penalizing general traffic flow. The significance of this work lies in its demonstration that reinforcement learning can be tailored to prioritize specific vehicle classes, such as emergency responders, within intelligent transportation systems. By integrating DVSL with DDPG, the study provides a viable method for real-time traffic adaptation that enhances safety and mobility in congested areas. The findings suggest that such systems can reduce response times for emergency services, potentially lowering fatality rates, while simultaneously improving overall traffic efficiency and reducing congestion at highway bottlenecks.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• emergency work zone conspicuity