Development of an Exoskeleton Rehabilitation Robot Framework for the Knee Joint Based on Predictive Assessment

Abstract

Common gait deviations in cerebral palsy can be divided into spastic hemiplegia and spastic diplegia gait patterns. These gait deviations require the use of a large amount of data obtained by clinical gait analysis through video, kinematic, kinetic, electromyographic sensors, and plantar pressure data to assess patient gait characteristics. In this paper, we use predictive forward dynamics to investigate the effects of the biarticular hamstrings (HAMS), gastrocnemius (GAS) and biceps femoris short head (BFSH) on the knee joint in gait cycle. We attempted to explore the characteristics for reproducing the patient's gait by modifying the parameters of model muscles. We applied mild, moderate and severe muscle weakness or contracture to the HAMS, GAS and BFSH muscle groups, respectively, and trained the model to walk at a self-selected speed, showing that with the more severe contracture, the non-swinging phase presented more severe knee hyperflexion and stronger knee torque, and the sensitivity for change is ranked by GAS>BFSH>HAMS. In swing phase, HAMS and GAS contractures aggravate the knee angle, whereas contractures of BFSH have a weak effect on knee angle. Mild HAMS muscle weakness accelerated walking speed, while moderate and severe HAMS muscle weakness hindered walking speed instead. BFSH muscle weakness is more sensitive to knee joint torque. Finally, a variable parameter impedance controller for the lower limb exoskeleton rehabilitation robot is developed. We apply the knee joint angle and torque parameters optimized by predictive forward dynamics simulation as the expected values for the robot to achieve customized tuning of the motion trajectory for the exoskeleton rehabilitation robot and meet the different rehabilitation stages.