CMA-MAPPO: Integrating Covariance Matrix Adaptation Evolution Strategy with Multi-Agent Proximal Policy Optimization for enhanced exploration in sparse-reward environments

Key Points

• Improved exploration in sparse-reward environments was achieved through the CMA-MAPPO method.
• Key evidence shows significant enhancement in performance metrics under specific setups.
• Theoretical model integrates covariance matrix adaptation and multi-agent proximal policy optimization for effective learning.
• This new method may enable better performance in complex multi-agent scenarios; however, further validation is necessary.