Kong, Yu (Chang'an University), Chen, Xinyu (Chang'an University), Mu, Chen (Chang'an University), Ning, Haijing (Chang'an University), Liu, Shumei (Chang'an University), An, Yisheng (Chang'an University)

Intersection Trajectory Planning: Modeling and Solution Algorithm Design Based on Petri Nets and Deep Reinforcement Learning

Scheduled for presentation during the Regular Session "S26a-Motion Planning, Trajectory Optimization, and Control for Autonomous Vehicles" (TH-LM-T26), Thursday, November 20, 2025, 10:30−10:50, Broadbeach 1&2

2025 IEEE 28th International Conference on Intelligent Transportation Systems (ITSC), November 18-21, 2025, Gold Coast, Australia

This information is tentative and subject to change. Compiled on October 18, 2025

Abstract

Efficient collaborative control of signal-free intersections is a major challenge faced by intelligent transportation systems, with the core issue being how to improve the safety and traffic efficiency of connected and automated vehicles (CAVs). To address the bottleneck of high model complexity and computational overhead in existing trajectory planning methods, this paper proposes an innovative framework, P2DT, which combines Petri net theory with deep reinforcement learning. First, a collaborative control model for intersections based on mixed-integer nonlinear programming (MINLP) is established, and a Petri net model based on intersection structure is designed, reformulating the resource allocation problem as a Markov Decision Process (MDP). Second, a Petri-Graph Neural Network (PGNN) architecture is proposed, developing a resource-operation relationship model and utilizing graph embedding techniques to capture the interaction features between vehicles. Additionally, the Deep Q-Networks (DQN) is employed for dynamic decision optimization. Numerical experiments show that the proposed method enables all automated vehicles to safely and efficiently cross the intersection, significantly improving computational efficiency while achieving satisfactory solutions.