An intelligent wheelchair control method based on POMDP user intention model

The model of user intention is realized based on partially observable Markov decision process (POMDP), and a control method of intelligent wheelchair based on the POMDP user intention model is presented, which considers the nondeterministic actions and the partial observability of the state. Based on the model of user intention and the navigation control action space, the optimal planning of the action is chosen by state prediction probability, which realizes the maximum of total value of user's expectation. First, the principle and the model of POMDP are introduced. Then the state space, the action space, the set of observations, the reward function and the probability of the POMDP user intention model of intelligent wheelchair are given. After that, the solution of POMDP user intention model based on probability density function is proposed. Finally the intelligent wheelchair navigation experiments based on the POMDP user intention model are carried out, and the experimental results verify the effectiveness and real-time property of the method.