Distributed PSO for dynamic intersection management: Enhancing traffic flow and safety in connected autonomous vehicles

Efficient intersection management remains a critical challenge for Connected and Autonomous Vehicles (CAVs), especially under dynamic traffic conditions that require balancing safety, throughput, and responsiveness. Existing cooperative protocols, such as virtual platooning and rule-based scheduling, offer decentralized control but typically rely on fixed synchronization points and static sequencing rules, limiting their adaptability in real-time environments. In this study, we propose PSO-DCPVP, a novel hybrid framework that integrates Particle Swarm Optimization (PSO) within the Distributed Clearing Policy for Virtual Platooning (DCPVP). The key innovation lies in dynamically computing mobile synchronization points for each vehicle based on local traffic states, enabling more flexible and context-aware platoon coordination. We evaluate the framework in a custom multi-agent simulation environment under three traffic scenarios: low, moderate, and high demand over a 600 s simulation horizon. Results demonstrate that PSO-DCPVP significantly increases intersection throughput, exceeding 2.1 pcu/s in congested settings while reducing average delay to below 0.03 s. Compared to baseline strategies such as FIFS and DCPVP, PSO-DCPVP demonstrates strong potential for real-world deployment in intelligent transportation systems.