Grey Reinforcement Learning for Incomplete Information Processing

Chunlin Chen, Daoyi Dong, Zonghai Chen

Abstract: New representation and computation mechanisms are key approaches for learning problems with incomplete information or in large probabilistic environments. In this paper, traditional reinforcement learning (RL) methods are combined with grey theory and a novel grey reinforcement learning (GRL) framework is proposed to solve complex problems with incomplete information. Typical example of mobile robot navigation is given out to evaluate the performance and practicability of GRL. Related issues are also briefly discussed.