Risk-aware Defensive Motion Planning for Distributed Connected Autonomous Vehicles

Decentralized connected autonomous vehicles that do not rely on a central controller for coordination and scheduling offer higher scalability and fault tolerance, making them suitable for large-scale deployments in dynamic traffic environments. Each vehicle serves as a network node, sharing uncertain information with other vehicles to achieve safe and comfortable decision-making and planning. In this paper, we propose a defensive motion planning method that takes into consideration the potential behaviors of surrounding vehicles to generate safe and comfortable trajectories. For risk assessment, we estimate future collision risks using two metrics: collision severity and collision probability. We employ a more accurate collision octagon as the integration region for numerical integration. To validate our approach, we construct a realistic-scale simulation environment that replicates actual traffic scenarios. Experimental results demonstrate that our method can effectively handle uncertain intentions and generate feasible, safe, and comfortable trajectories.