CCL: Collaborative Curriculum Learning for Sparse-Reward Multi-Agent Reinforcement Learning via Co-evolutionary Task Evolution
This addresses the problem of delayed and shared feedback in multi-agent systems for researchers, but it appears incremental as it builds on existing curriculum learning approaches.
The paper tackles the challenge of sparse rewards in multi-agent reinforcement learning by proposing a curriculum learning framework that refines tasks, generates subtasks, and co-evolves agents with the environment, resulting in outperformance over existing methods in cooperative tasks.
Sparse reward environments pose significant challenges in reinforcement learning, especially within multi-agent systems (MAS) where feedback is delayed and shared across agents, leading to suboptimal learning. We propose Collaborative Multi-dimensional Course Learning (CCL), a novel curriculum learning framework that addresses this by (1) refining intermediate tasks for individual agents, (2) using a variational evolutionary algorithm to generate informative subtasks, and (3) co-evolving agents with their environment to enhance training stability. Experiments on five cooperative tasks in the MPE and Hide-and-Seek environments show that CCL outperforms existing methods in sparse reward settings.