Abstract: Multi-agent reinforcement learning has emerged as a promising approach for the control of multi-robot systems. Nevertheless, the low sample efficiency of MARL poses a significant obstacle to ...