抄録
Operating an electric wheelchair indoors in confined spaces requires considerable skill. This paper presents an obstacle avoiding support system for an electric wheerchair, using reinforcement learning. The system adds virtual repulsion from the obstacles to manual input from the user. The virtual repulsion is tuned corresponding to each unique environment or user condition. Then, the size of the virtual repulsions are determined by the reinforcement learning controller. The newly proposed scheme is numerically evaluated on a simulation example. In the future, further experiments using an electric wheelchair will be performed. Also, the proposed algorithm will be installed on the DSP board.
