Yize Mi

Unveiling Complex Collective Behaviors from Simple Rewards

Yize Mi, Jianan Li, Liang Li, Shiyu Zhao

Multi-agent Reinforcement Learning (MARL) holds great potential for robot swarms, but the black-box nature of neural policies complicates strategic analysis, limiting multi-robot applications. Furthermore, complex swarm behaviors can surprisingly emerge from simple rewards withou…Read more