Sampling-Based Coordination-Informed Multi-Objective Multi-Robot Reinforcement Learning
For multi-robot systems needing to balance competing objectives under distributed constraints, this framework provides a decentralized solution with strong empirical gains.
This work introduces CIMORL, a framework for multi-objective multi-robot reinforcement learning that achieves a 21.2% hypervolume improvement over state-of-the-art baselines in cooperative and adversarial scenarios, validated with real-world Crazyflie drone experiments.
Multi-robot systems must simultaneously optimize competing objectives while maintaining coordinated behavior. Existing multi-agent reinforcement learning approaches often rely on fixed or centralized coordination, which limits adaptability and violates distributed constraints. This work introduces the Coordination-Informed Multi-Objective Reinforcement Learning (CIMORL) framework, integrating a distributed weight prediction mechanism, a privileged expert training strategy, and theoretical guarantees for Pareto-optimal solutions. We present the base CIMORL method alongside two sampling-based variants, CIMORL-TS (Tree Search) and CIMORL-MPPI (MPPI), which leverage privileged global information during training to enable fully decentralized deployment. Experimental validation in cooperative and adversarial scenarios demonstrates a $21.2\%$ hypervolume improvement and superior policy stability compared to state-of-the-art baselines. Real-world experiments with Crazyflie drones further validate the framework's robustness in resource allocation and multi-attacker multi-defend scenarios under partial observability.