Regularized Soft Actor-Critic for Behavior Transfer Learning
This addresses the need for efficient methods to partially imitate behaviors while completing tasks, which is incremental as it builds on existing Soft Actor-Critic and imitation learning approaches.
The paper tackles the problem of agents imitating demonstrated behaviors that may conflict with task objectives by proposing Regularized Soft Actor-Critic, which formulates imitation as a constraint under a Constrained Markov Decision Process framework, and evaluates it on continuous control tasks for video games.
Existing imitation learning methods mainly focus on making an agent effectively mimic a demonstrated behavior, but do not address the potential contradiction between the behavior style and the objective of a task. There is a general lack of efficient methods that allow an agent to partially imitate a demonstrated behavior to varying degrees, while completing the main objective of a task. In this paper we propose a method called Regularized Soft Actor-Critic which formulates the main task and the imitation task under the Constrained Markov Decision Process framework (CMDP). The main task is defined as the maximum entropy objective used in Soft Actor-Critic (SAC) and the imitation task is defined as a constraint. We evaluate our method on continuous control tasks relevant to video games applications.