Personalized passive training control strategy for a lower limb rehabilitation robot with specified step lengths

Passive training is a crucial rehabilitation strategy in lower limb rehabilitation robotics. However, most rehabilitation robots currently use a fixed training mode, which cannot be personalized according to the individual characteristics and rehabilitation needs of patients. To increase training effectiveness and improve the rehabilitation experience, this study introduces a personalized passive training control strategy for a rehabilitation robot, focusing on gait training in early-stage rehabilitation for patients with lower limb dysfunction. First, a personalized gait trajectory generation model is developed based on radial basis function neural networks (RBFNN). This model aims to generate personalized gait trajectories to accommodate different patients with varying step lengths. A proportional-integral-derivative (PID) controller based on the motor speed loop is subsequently implemented to enable the robot to accurately track the generated personalized gait trajectories. To evaluate the effectiveness of the control strategy, five healthy participants were recruited for experimental verification. The research findings demonstrate that the model based on the RBFNN can generate gait patterns with specified step lengths with high precision and accuracy. Compared with the previous state-of-the-art Gaussian process regression (GPR) method, this model reduces the prediction error of lower limb joint angles and center of gravity (CG) trajectories by 11.34% and 6.49%, respectively. Moreover, the control performance of the PID controller is verified through experimental comparisons and analysis of literature results. Therefore, this control strategy can effectively coordinate the control of the ankle, knee, and hip joints, as well as the CG, and achieve personalized gait rehabilitation training for specific step lengths.