Application of Attention-Based Multi-Agent Deep Reinforcement Learning in Team Sports Collaboration

With the widespread application of deep reinforcement learning in multi-agent systems, improving collaboration efficiency and strategy execution among agents has become an urgent challenge. This paper proposes an attentionbased multi-agent deep reinforcement learning algorithm-Attention-MADDPG-to enhance cooperation efficiency in team movement tasks. By introducing an attention mechanism, the method dynamically adjusts the degree of mutual focus among agents, effectively optimizing strategy execution. Experiments conducted in multiple football teams from Sichuan demonstrate that the Attention-MADDPG algorithm exhibits faster convergence during training and significantly outperforms traditional MADDPG and DQN algorithms in collaboration efficiency and strategy execution. The experimental results validate the effectiveness of the proposed method, providing new insights and approaches for optimizing cooperation in multi-agent systems.