On-line NNAC for a Balancing Two-Wheeled Robot Using Feedback-Error-Learning on the Neurophysiological Mechanism

A biomimic control structure is designed based on the sensorimotor system and the central nervous system of human brain. The main control center is focus on the cerebellar cortex functioning in supervised training. And then a method based on cerebellar model with feedbackerror-learning on the neurophysiological mechanism was proposed to control a balancing two-wheeled robot which is a bionic self-balancing plant imitating human body. The cerebellar cortex as a forward controller was an on-line neural network adaptive controller and reflected the cerebellar function in balancing control problem. From the simulation including balancing control experiment, disturb experiment and mass changed experiment, we can see that the robot can be balanced in fixed position well by the biomimic control structure, and get a better result comparing the traditional feedback control.