Provably Efficient Algorithms for Multi-Objective Competitive RL

                  Tiancheng Yu 1 Yi Tian 1 Jingzhao Zhang 1 Suvrit Sra 1

              Abstract               average return to a target set small as long as this set satisfies
                                 a condition called approachability (Blackwell, 1956).
    We study multi-objective reinforcement learning
    (RL) where an agent’s reward is represented as      The approachability theorem applies to multi-objective
    a  ...