Multi-objective Trajectory Planning Based on Integrated NSGA-III for Isokinetic Rehabilitation

Upper limb exoskeletons can autonomously help patients after stroke regain the function of upper limb. To enhance the safety and comfort of rehabilitation training, a multi-objective trajectory optimization method specifically tailored for isokinetic rehabilitation is proposed. Initially, original trajectory for isokinetic training is simulated using piecewise polynomials. Subsequently, objective functions are constructed based on trajectory smoothness, duration, and energy consumption, with energy consumption derived from joint torque and joint angular velocity. Finally, we utilize the integrated non-dominated sorting genetic algorithm III (INSGA-III) with non-uniform mutation and guided crossover to optimize the motion trajectory, conforming to human motion patterns. Experimental results demonstrate the superior convergence capability of INSGA-III compared to NSGA-III. Meanwhile, the trajectory angles, velocities, and accelerations remain continuous and smooth, with no noticeable abrupt changes in impact curves, which indicates that the proposed trajectory effectively fulfills the requirements of rehabilitation training.