Distributed Nonconvex Optimization and Application to UAV Optimal Rendezvous Formation

Abstract

A distributed multiagent deep reinforcement learning algorithm (DMADRLA) with theoretical guarantees is proposed for the distributed nonconvex constraint optimization problem. This algorithm provides an innovative theoretical framework for distributed nonconvex optimization problems (DNCOPs) by combining traditional distributed constraint optimization and multiagent deep reinforcement learning methods. This combination eliminates the need for general assumptions on the cost function, enabling a more comprehensive view of distributed nonconvex optimization strategies. It allows for the analysis of both traditional distributed constrained optimization and multiagent deep reinforcement learning methods in one unified approach. Finally, the effectiveness of the algorithm is verified through numerical simulations and experimental verification.