Multi-agent Path Finding for Mixed Autonomy Traffic Coordination

TL;DR

BK-PBS tackles mixed autonomy highway coordination by integrating an offline conditional HDV behavior predictor into a PBS planner, with a low-level multi-phase A* that respects kinematics via motion primitives. The approach enables proactive coordination by branching on priority and conditioning HDV trajectories on CAV actions, learned via a Graph Convolutional Network; evaluated on highway merging scenarios across varying CAV penetration $\alpha$ and traffic density $\lambda$, it reduces controllable collision rates and total travel delay compared with baselines (BK-M-A*, IDM+MOBIL, RL-PPO). Key findings include that BK-PBS achieves robust safety and improved delay performance in dense traffic, particularly when comparing at equal collision rates where it can reduce travel delay by 15–20% relative to IDM+MOBIL. The work advances mixed-traffic planning by demonstrating how HDV behavior prediction can be embedded into MAPF-style coordination, with implications for broader multi-human multi-robot coordination tasks.

Abstract

In the evolving landscape of urban mobility, the prospective integration of Connected and Automated Vehicles (CAVs) with Human-Driven Vehicles (HDVs) presents a complex array of challenges and opportunities for autonomous driving systems. While recent advancements in robotics have yielded Multi-Agent Path Finding (MAPF) algorithms tailored for agent coordination task characterized by simplified kinematics and complete control over agent behaviors, these solutions are inapplicable in mixed-traffic environments where uncontrollable HDVs must coexist and interact with CAVs. Addressing this gap, we propose the Behavior Prediction Kinematic Priority Based Search (BK-PBS), which leverages an offline-trained conditional prediction model to forecast HDV responses to CAV maneuvers, integrating these insights into a Priority Based Search (PBS) where the A* search proceeds over motion primitives to accommodate kinematic constraints. We compare BK-PBS with CAV planning algorithms derived by rule-based car-following models, and reinforcement learning. Through comprehensive simulation on a highway merging scenario across diverse scenarios of CAV penetration rate and traffic density, BK-PBS outperforms these baselines in reducing collision rates and enhancing system-level travel delay. Our work is directly applicable to many scenarios of multi-human multi-robot coordination.

Figures (3)

• Figure 1: Mixed traffic coordination: 1) The left portion describes a collision between a CAV $i^c$ and a HDV $j^h$. To resolve this CAV-HDV collision, BK-PBS introduces priority between these two vehicles, and replan the CAV trajectory using M-A* while re-predict the HDV trajectory using the conditional prediction model $b_{\theta}$. 2) The right portion describes a collision between a CAV $m^c$ and a CAV $n^c$. To resolve this CAV-CAV collision, BK-PBS introduces priority between these two vehicles, and replan each of the CAV trajectory using M-A*.
• Figure 2: Highway Simulator: CAVs are colored in green and HDVs are colored in blue
• Figure 3: Controllable collision rate under different vehicle arrival rates and CAV penetration rates