The Max-Min Formulation of Multi-Objective Reinforcement Learning: From Theory to a Model-Free Algorithm

TL;DR

This paper introduces a max-min framework for multi-objective reinforcement learning, providing new theoretical insights and a practical model-free algorithm that improves performance in balancing multiple goals.

Contribution

It presents a novel max-min formulation for multi-objective reinforcement learning, along with a new theory and a practical algorithm that enhances fairness and performance.

Findings

Theoretical advancement in multi-objective RL with max-min approach.

Proposed algorithm outperforms baseline methods.

Framework emphasizes fairness among multiple objectives.

Abstract

In this paper, we consider multi-objective reinforcement learning, which arises in many real-world problems with multiple optimization goals. We approach the problem with a max-min framework focusing on fairness among the multiple goals and develop a relevant theory and a practical model-free algorithm under the max-min framework. The developed theory provides a theoretical advance in multi-objective reinforcement learning, and the proposed algorithm demonstrates a notable performance improvement over existing baseline methods.