R2BC: Multi-Agent Imitation Learning from Single-Agent Demonstrations
Researchers have developed Round-Robin Behavior Cloning (R2BC), a novel method for training multi-robot systems using sequential, single-agent demonstrations. This approach allows a single human operator to teach complex multi-agent behaviors by teleoperating one robot at a time, without needing synchronized multi-agent demonstrations. R2BC has shown performance comparable to or exceeding traditional behavior cloning methods in simulations and has been successfully deployed on physical robots. AI
IMPACT Enables more efficient training of multi-robot systems, potentially accelerating adoption in complex robotic tasks.