A Multi-Agent system for Multi-Objective constrained optimization

The paper introduces MAMO (Multi-Agent system for Multi-Objective constrained optimization), which utilizes multi-agent reinforcement learning (RL) to address cost-minimization problems with performance constraints in dynamic environments. MAMO innovatively decouples task execution from objective design by framing the selection of reward weights as a learning problem, thus enhancing the autonomy and robustness of RL solutions for constrained optimization. This approach is significant for practitioners as it allows for adaptive trade-offs between primary objectives and constraint violations, improving performance in non-stationary settings.