Dynamic Modeling and Compliant Control for a Lower Extremity Exoskeleton Robot Based on BP Neural Network

In order to realize the active control of the swinging leg of a 2-DOF lower extremity exoskeleton robot with complex multi-link drive, a dynamic modeling and compliant control method based on BP neural network is proposed in this paper. Firstly, the dynamics data of the joints are acquired by static experiments under no-load and the double-joint motion experiments under sine and cosine signals. Then, the dynamic model of the robot is established by building and training a BP neural network. A variable frequency motion experiment under no-load is taken to verify the correctness of the model. Based on the BP neural network and the PID controller, a compliant control method is designed. Finally, no-load static experiment and motion tracking experiment are carried out. Experiments show that the trained model can well estimate the human-robot interaction torque under both static and dynamic conditions, and motion intention recognition and motion tracking are realized by the designed control method. In addition, the weight of the exoskeleton is compensated to reduce the burden of the exoskeleton on people.