A case-based reinforcement learning for probe robot path planning

摘要

This paper discusses the application of case-based learning for probe robot path planning in unknown environments. Case-based learning which makes use of past experience, is an incremental learning process. This paper proposes an algorithm of introducing reinforcement learning to case-based-reasoning, which makes full use of knowledge acquired by reinforcement learning to construct and extend the case-library. This method can enhance the adaptability of robot to unknown environments and solve the problem of case acquiring as well as poor real-time performance, high learning risk of reinforcement learning. Also, with the forget-rule, case-library can be updated in time so that efficiency of case searching and learning is increased. As the learning progressing and the case-library dynamically updated, robot's intelligence has been greatly improved. A simulation shows the validity and feasibility of this method.