Soft Actor-Criticの改良による出力抑制と頑健化

抄録

Recently, soft actor-critic (SAC) is employed for robot control. Although its ability to maximize policy entropy is expected to achieve robustness to noise and perturbation in robot control, the priority of maximizing the policy entropy is automated based on equality constraint to lower bound. Therefore, sufficient robustness is no longer expected. To resolve this issue, this paper proposes a new automation method on SAC using a slack variable for handling inequality constraint. As a result, the modified SAC achieved the higher robustness. In addition, as a side effect, it successfully suppressed its outputs.