Dependency of Parameter Values in Reinforcement Learning for Navigation of a Mobile Robot on the Environment

Reinforcement learning is suitable for navigation of a mobile robot due to its learning ability without supervised information. Reinforcement learning, however, has difficulties. One is its slow learning, and the other is the necessity of specifying its parameter values without prior information. We proposed to introduce sensory signals into reinforcement learning to improve its learning performance, and to optimize its parameter values in reinforcement learning by a genetic algorithm with inheritance. The latter has to specify the parameter values for every new environment, which is impractical due to huge computational time. In this paper, we propose to analyze the dependency and sensitivity of the values of parameters on the environment for predicting the values of parameters for a novel environment without optimization. Computer experiments clarify the dependency of the values of parameters on the environment and their sensitivities. Keywords—reinforcement learning, genetic algorithm, navigation of a mobile robot, parameter dependency