A digital twin-driven trajectory tracking control method of a lower-limb exoskeleton

A direct drive motor is the main component for tracking the gait rehabilitation training trajectory of a lower-limb exoskeleton (LLE). Aiming at the movement instability caused by the changes in the moment of inertia and assembly characteristics of the LLE, a trajectory tracking control method based on the digital twin model is proposed in the study. Firstly, the key characteristic parameters of LLE driven by the direct drive motor are extracted to establish a virtual twin model of LLE. Secondly, the digital twin model is driven by the physical-entity state data and the control parameters of the motor servo are optimized through the twin model based on an adaptive feedback control strategy. Finally, in order to improve the real-time feedback accuracy between the twin model and the physical entity, with the depth deterministic policy gradient and particle swarm optimization algorithm (DDPG-PSO), the parameter matching error between the twin model and the physical entity is reduced. In this way, a digital twin-driven compound control algorithm is obtained. In addition, the proposed method was verified through three sets of experiments. In Experiment 1, the virtual joint angle trajectory θ f i c of the twin model was compared with the actual joint angle trajectory θ a c t and the average error was no more than 0.05, indicating that the twin model could accurately restore the motion trajectory of the physical entity. In Experiment 2, by comparing with tracking effects of Model-free adaptive control, the adaptive feedback control of digital twin-driven has better stability, and can effectively resist external interference. In Experiment 3, under the no-load condition, the algorithm converged to the optimal solution after 40 iterations. In addition, dynamic parameter changes could be detected in real time, thus proving the rapid convergence and good performance of the DDPG-PSO algorithm. • Restore the physical entity in virtual space and map its motion state. • Establishes a digital twin model of LLE. • With the twin model, the external disturbance during the movement of the physical entity is estimated. • Proposes an adaptive feedback control strategy driven by digital twins. • Use the DDPG-PSO algorithm to optimize the digital twin model.