The motion control of lower extremity exoskeleton based on RBF neural network identification

A lower extremity exoskeleton is designed and a novel control method based on RBF neural network identification is proposed to guarantee the exoskeleton identifies the pilot's motion intention independently. The propose method considers the pilot wearing assistance exoskeleton as an unknown system, and the RBF neural network is employed to identify the system's motion trajectories and the identification results are used as the exoskeleton's desired trajectories. The network's error function is obtained indirectly by the weighted summation of exoskeleton's tracking error and the interaction torque between pilot and exoskeleton. A human-machine system model including the elastic interaction torque is established according to the structural features of the designed one DOF exoskeleton, and then simulation is conducted based on it. The simulation results show the propose control method can identify the pilot's motion intention well and reduce the exoskeleton's drag torque acting on pilot dramatically. When dealing with the measurement noises, this method also shows good robustness. In addition, the exoskeleton can give pilot different tactile feedback through regulating the interaction torque's scale factor.