Toward Verifiably Safe and Efficient Autonomous Intersections: A 6G-Assisted Transformer and TPN Cooperative Framework

This paper presents a formal framework for cooperative and collision-free scheduling of autonomous vehicles at signal-free intersections. The core of our approach leverages Timed Petri Nets (TPN) to model intersection traffic as a verifiable resource allocation system, ensuring strict safety guarantees through spatial-temporal partitioning. A key innovation is the integration of a transformer-based prediction module that captures complex vehicle interactions, enabling accurate passing-order decisions. Combined with adaptive model predictive control (MPC), this transformer-enhanced strategy coordinates multiple intersections in real-time. Simulation results demonstrate an 18~22% reduction in congestion delays compared to conventional methods. Designed for 6G-enabled Intelligent Internet of Things (IIoT) infrastructures, the framework supports ultra-reliable low-latency communication and edge intelligence, providing a robust solution for next-generation transportation networks.