Dynamic Learning from Neural PID Output-Feedback Control for a Lower Limb Exoskeleton Rehabilitation Robot

With the intensification of population aging and the increasing number of patients with lower limb dysfunction, the research and precise control of lower limb rehabilitation robots have become increasingly important. In this study, a sitting and lying lower limb exoskeleton robot model is developed based on the CoppeliaSim platform. For the nonlinear system of a two-degree-of-freedom single-leg robot with uncertainty, an improved PID control method based on deterministic learning is proposed. Under the condition of persistent excitation, this control method accurately identifies the unknown dynamic characteristics of the system by introducing a neural network, ensuring precise control of lower limb movements. Learning-based PID controllers are constructed using the gained knowledge to achieve fast, smooth tracking and greatly improve the safety of rehabilitation control. The results of simulation experiments show that the proposed control strategy in this paper is effective.