An incremental behavior learning based on reinforcement learning with schema extraction mechanism for autonomous mobile robot

Recently, a number of skillful robots have been developed. However it can so far only demonstrate preprogrammed motions according to external stimuli. In contrast, humans can learn new motions such as catching a ball, in spite of his/her high dimensional sensorimotor DOF. In this learning process, it can be hypothesized that the learner actively constrains the DOF by him/her-self using learning skills, in this paper referred to as schema. In this study, a learning method for autonomous mobile robots operating in unknown environments is proposed, where not only a learning mechanism for sensorimotor mappings but also an extraction/re-use mechanism of the schemata (i.e. constraint rules for learning) is implemented. Through the results of simulations and real experiments of mobile robot navigation, the validity of the proposed method is clarified.