Multi-Robot Collaboration for Dynamic Environments: A Study Based on the Improved PPO Algorithm

With the expanding applications of multi-robot systems in complex dynamic environments, efficiently coordinating multiple robots to complete intricate tasks has become a hot research topic. This study proposes an improved Proximal Policy Optimization (PPO) algorithm, which optimizes inter-robot collaboration through reinforcement learning to enhance the adaptability and stability of multi-robot systems in dynamic environments. We employ a Markov Decision Process (MDP) framework and coordinate each robot using a joint reward function to achieve global task optimization in complex tasks. Experimental results demonstrate that the improved PPO significantly outperforms traditional PPO and MPC methods in terms of task completion efficiency, collaboration efficiency, and stability, especially when facing environmental disturbances and increased task complexity. The improved PPO algorithm shows stronger adaptability and robustness, proving its superiority in multi-robot collaborative tasks.