Science

Structural bias enables reliable sim-to-real transfer in reinforcement learning for robot swarms

Constraining policies to select behavior modules allows reinforcement learning transfer from simulation to real robot swarms.