Continuous Control for Searching and Planning with a Learned Model

• DOI: 10.48550/arxiv.2006.07430
• Publication Date: 2020-01-01
• Authors: Xuxi Yang; Werner Duvaud; Peng Wei
• Keywords: 0301 basic medicine; FOS: Computer and information sciences; Computer Science - Machine Learning; 03 medical and health sciences; Artificial Intelligence (cs.AI); Computer Science - Artificial Intelligence; Machine Learning (cs.LG)

Decision-making agents with planning capabilities have achieved huge success in the challenging domain like Chess, Shogi, and Go. In an effort to generalize ability more general tasks where environment dynamics are not available agent, researchers proposed MuZero algorithm that can learn dynamical model through interactions environment. this paper, we provide a way necessary theoretical results extend generalized environments continuous action space. Through numerical on two relatively low-dimensional MuJoCo environments, show outperforms soft actor-critic (SAC) algorithm, state-of-the-art model-free deep reinforcement learning algorithm.