Sequential Distributed Optimization for Cooperative Collision Avoidance via Agent-Wise CCP-PSM

초록

This paper proposes Agent-wise CCP-PSM, a distributed optimization framework for collision avoidance in high-density multi-agent systems. The proposed method linearizes non-convex distance constraints using the Convex-Concave Procedure (CCP) and solves each agent's local problem via a computationally efficient Projected Subgradient Method (PSM). By incorporating collision avoidance as soft penalties, the framework improves numerical robustness and maintains solution availability in congested scenarios. Furthermore, a Gauss-Seidel-based sequential update enables each agent to utilize the most recent information from others while keeping the subproblem dimension independent of the total number of agents, improving coordination efficiency. Within a receding-horizon framework, the method generates collision-avoidance trajectories in real time while empirically reducing minimum-separation violations. Its computational complexity with respect to the number of agents and planning horizon is analyzed and validated through simulations. Results demonstrate that the proposed framework provides a favorable balance among minimum-separation performance, computational efficiency, and scalability across diverse scenarios.

키워드