A dynamic reinforcement learning strategy was introduced to establish the correlation between individual operating habits and collision risk based on a triple-reward system in order to resolve the impact of individual operating habits on the shared motion control of intelligent wheelchair robots (WR). A fuzzy reinforcement learning state fusion-based shared control strategy was proposed,which could adapt to user behavior while ensuring safety. A distance fuzzy reasoning algorithm was employed to develop a direction intention recognition model based on seat pressure,which served as the foundation for establishing a human-machine shared control framework in order to achieve intelligent robot control. The current and predictive reward functions were established via the Gaussian function and deviation rate,respectively,focusing on the deviation between the user's intended direction and the robot's actual direction in order to estimate user operating habits. A task reward function was created according to boundary distance in order to predict human-machine safety. The correlation between user operating habits and safety was constructed by utilizing the fuzzy reinforcement learning strategy and the triple-reward system in order to dynamically adjust the user control weight within the shared control to adapt to individual habits. Then the precision and safety of human-machine shared control were enhanced. The effectiveness of the proposed algorithm was verified by experiments in a test environment.
万方数据
