Carrying robot system control based on neural network

The virtual torque control is a relative effective control law for the lower extremity carrying exoskeleton robot, which needs no direct measurements from the pilot or the human-machine interface, but it needs the exact mass properties and its PD controller is not applicable for the serious nonlinear system. To overcome these defaults, the virtual prototyping software ADAMS is introduced to model the exoskeleton and the model is also used to the neural network (NN) dynamic mathematics model study. At the same time, the NN is used to the dynamic feed-forward controller to compensate the human-machine interaction force. Theoretical analyse and simulation results test the feasibility and validity of this control method.