To improve the mode and precision of wheelchair/nursing-bed automatic docking, a novel central embedded wheelchair/nursing-bed automatic docking method based on grid map is proposed. Firstly, Laplace operator and Iterative Closest Point (ICP) algorithm are used to filter and match Lidar point cloud, and the linear features of V-shaped artificial landmark are fitted by Split-merge method and least square method. Then Extended Kalman Filter (EKF) is used to fuse Inertial Measurement Unit (IMU) and odometer data to realize the localization of the bed and wheelchair. Meanwhile, the grid map is used for path planning. Based on the center-line of the two rear wheels and the angular bisector of V-shaped artificial landmark, the wheelchair pose is adjusted in real-time to ensure that the wheelchair gradually approaches the bed along the angular bisector of V-shaped artificial landmark. The yaw angle is reduced by using the improved Proportion Integration Differentiation (PID). 9 sets of experimental data, ie. (x, y,  ) were collected at different starting positions during the docking process. The results show that the yaw angle of the wheelchair during the docking process is controlled within 2.5°, and the distance deviation between the final position and the ideal position of the wheelchair after docking is controlled within 0.02m. In the case of light interference with different luminous fluxes, the docking can still maintain good performance. The proposed docking algorithm has the robust performance of rapid response and low steady error, which can greatly improves the self-care ability of the bedridden elderly, and reduces the labor intensity of the nursing staff.

BACKGROUND: During neurological rehabilitation training for patients with lower limb dysfunction, active rehabilitation training based on interactive force recognition can effectively improve participation and efficiency in rehabilitation training. OBJECTIVE: This study proposes an active training strategy for lower-limb rehabilitation robots based on a spring damping model. METHODS: The active training strategy included a kinetic model of the human-machine system, calculated and verified using a pull-pressure force sensor We used a dynamic model of the human-machine system and tensile force sensors to identify the human-machine interaction forces exerted by the patient Finally, the spring damping model is used to convert the active interaction force into the offset angle of each joint, obtaining the active interaction force followed by the active movement of the lower limbs RESULTS: The experimental results showed that the rehabilitation robot could follow the active interaction force of the subject to provide assistance, thus generating the following movement and effectively helping patients improve joint mobility. CONCLUSION: The active flexibility training control strategy based on the virtual spring damping model proposed in this study is feasible, and motion is stable for patients with lower limb dysfunction after stroke Finally, the proposed active training method can be implemented in future work in other rehabilitation equipment and combined virtual reality technology to improve rehabilitation training experience and increase patient participation.