Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Analyzing Collision Rates in Large-Scale Mixed Traffic Control via Multi-Agent Reinforcement Learning

Muyang Fan

TL;DR

This study tackles safety in large-scale mixed traffic control governed by MARL by examining collision-rate determinants. It introduces a decentralized Dec-POMDP-based MARL framework for RVs, integrating HV IDM dynamics and mixed signalized/unsignalized intersections, optimized with Rainbow-DQN. Through network-scale experiments on a 14-intersection SUMO network, it reveals how intersection mix, traffic density, and turning strategies influence collision risk. The findings provide actionable guidance for designing safer MARL-based mixed traffic control systems and highlight conditions under which safety can be robustly improved.

Abstract

Vehicle collisions remain a major challenge in large-scale mixed traffic systems, especially when human-driven vehicles (HVs) and robotic vehicles (RVs) interact under dynamic and uncertain conditions. Although Multi-Agent Reinforcement Learning (MARL) offers promising capabilities for traffic signal control, ensuring safety in such environments remains difficult. As a direct indicator of traffic risk, the collision rate must be well understood and incorporated into traffic control design. This study investigates the primary factors influencing collision rates in a MARL-governed Mixed Traffic Control (MTC) network. We examine three dimensions: total vehicle count, signalized versus unsignalized intersection configurations, and turning-movement strategies. Through controlled simulation experiments, we evaluate how each factor affects collision likelihood. The results show that collision rates are sensitive to traffic density, the level of signal coordination, and turning-control design. These findings provide practical insights for improving the safety and robustness of MARL-based mixed traffic control systems, supporting the development of intelligent transportation systems in which both efficiency and safety are jointly optimized.

Analyzing Collision Rates in Large-Scale Mixed Traffic Control via Multi-Agent Reinforcement Learning

TL;DR

This study tackles safety in large-scale mixed traffic control governed by MARL by examining collision-rate determinants. It introduces a decentralized Dec-POMDP-based MARL framework for RVs, integrating HV IDM dynamics and mixed signalized/unsignalized intersections, optimized with Rainbow-DQN. Through network-scale experiments on a 14-intersection SUMO network, it reveals how intersection mix, traffic density, and turning strategies influence collision risk. The findings provide actionable guidance for designing safer MARL-based mixed traffic control systems and highlight conditions under which safety can be robustly improved.

Abstract

Vehicle collisions remain a major challenge in large-scale mixed traffic systems, especially when human-driven vehicles (HVs) and robotic vehicles (RVs) interact under dynamic and uncertain conditions. Although Multi-Agent Reinforcement Learning (MARL) offers promising capabilities for traffic signal control, ensuring safety in such environments remains difficult. As a direct indicator of traffic risk, the collision rate must be well understood and incorporated into traffic control design. This study investigates the primary factors influencing collision rates in a MARL-governed Mixed Traffic Control (MTC) network. We examine three dimensions: total vehicle count, signalized versus unsignalized intersection configurations, and turning-movement strategies. Through controlled simulation experiments, we evaluate how each factor affects collision likelihood. The results show that collision rates are sensitive to traffic density, the level of signal coordination, and turning-control design. These findings provide practical insights for improving the safety and robustness of MARL-based mixed traffic control systems, supporting the development of intelligent transportation systems in which both efficiency and safety are jointly optimized.

Paper Structure

This paper contains 11 sections, 7 equations, 3 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Methodology
  4. Reinforcement Learning Formulation
  5. Modeling of Human-driven Vehicles
  6. Integrated Mixed Control of Intersections
  7. Evaluation Metrics
  8. Experiments and Results
  9. Exeperiment Set-up
  10. Results
  11. Conclusion and future work

Figures (3)

  • Figure 1: The collision rate under different configurations: Collision rates under different signal and unsignal intersection configurations. As the number of RV-controlled unsignal intersections increases, the overall collision rate exhibits a decreasing trend. This indicates that RV-based control strategies are more effective than traditional traffic signal control in enhancing safety performance within large-scale mixed traffic networks.
  • Figure 2: The collision rate under different vehicle numbers: Reducing the overall vehicle volume leads to a noticeable decrease in collision rates in large-scale mixed traffic control (MTC) networks managed by MARL. This reduction effect is more significant in configurations with fewer RV-controlled unsignal intersections.
  • Figure 3: The collision rate under different turning movements: With all left-turn movements converted into straight-through movements, most configurations exhibit a noticeable decrease in collision rates, indicating improved safety performance. However, the 10U+4S configuration shows an opposite trend, where collision rates increase across multiple RV penetration levels, suggesting that the removal of left-turn phases may cause increased conflict interaction under this specific network topology.