MADRL-based Intelligent Cooperative Formation Control Algorithm for UAV Swarms

To address the challenges of high robustness and the complexities of coordinated control in multi-UAV formation, we developed a distributed UAV swarm intelligent collaborative formation control algorithm based on MAPPO (MAPPO-DSFC), drawing inspiration from entralized Training with Decentralized Execution (CTDE) and multi-agent deep reinforcement learning (MADRL). Each UAV makes real-time decisions by sensing information about neighboring UAVs and obstacle with constrained communication radius. Simulation results show that the proposed algorithm exhibits superior performance in flight efficiency, formation stability, and formation reconstruction compared to other existing algorithms.