Sampling-Based Coordination-Informed Multiobjective Multirobot Reinforcement Learning

doi:10.1109/tro.2026.3723950

Sampling-Based Coordination-Informed Multiobjective Multirobot Reinforcement Learning

2026-08-13

0

PRE

AI

Paolo Robuffo Giordano

C

Claudio Pacchierotti

DOI：10.1109/tro.2026.3723950

Original

Original request for help

Save

Abstract

En 中文

Multirobot systems must simultaneously optimize competing objectives while maintaining coordinated behavior. Existing multiagent reinforcement learning approaches often rely on fixed or centralized coordination, which limits adaptability and violates distributed constraints. This work introduces the coordination-informed multiobjective 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-tree search (TS) and CIMORL-model predictive path integral (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 multiattacker multidefend scenarios under partial observability.

Keywords:

Distributed control

graph neural network

multiobjective optimization

reinforcement learning

Sampling-Based Coordination-Informed Multiobjective Multirobot Reinforcement Learning

Abstract

Journal

Organization

Cited Papers