Structural design and intelligent control method of rehabilitation exoskeleton robot

With the global aging and increasing disabled population, the patients with physical disability are facing with great challenges in improving life quality. As an assisting technology, rehabilitation exoskeleton robot helps patients to recover limb functions and improve life quality through rehabilitation training. In this study, a multimode information perception model based on man-machine coupling dynamics was proposed. It predicts and controls motion trails by combining with LSTM neural network. This method increases flexibility, adaptability and control accuracy of exoskeleton, improves the rehabilitation training efficiency and decreases dependences on human power. A multi-joint exoskeleton rehabilitation robot arm conforming to human sports anatomy was designed and kinematics analysis was carried out with SolidWorks modeling through D-H parametric method. In addition, a multimode information perception system of upper limb exoskeleton was designed, which realized more accurate recognition of motion intention and track prediction by using the multimode information fusion strategy. According to experimental results, the rail accuracy can be increased significantly by combining cubic polynomial interpolation algorithm and LSTM neural network, and the error decreases from 3.2°/s to 1.3°/s. The control effect of exoskeleton rehabilitation robotic arm is optimized.