Design and implementation of a kind of neural networks robustness controller for variable structure bicycle robot's track-stand motion

Abstract

This paper focus on dynamic modeling and a kind of controller for the variable structure bicycle robot's (which called VSBR) track-stand motion when its front wheel is fixed at 45 degrees about its front fork. Firstly, making dynamic analysis for the VSBR whose front wheel is fixed at arbitrary degrees about its front fork. The roll angle, yaw angle, the angle front fork rotated and the angle front wheel rolling are chosen as generalized coordinates. A kind of dynamic model of VSBR is built based on Routh equation. The angle of front fork rotated is fixed at 45 degrees in this dynamic model. And the angle of back wheel rolling is 0 in the track-stand motion. At last two dynamic differential equations are presented and the dynamic model is built. The robust controller is designed based on RBF neutral network arithmetic. Feedforward term in the controller is used for nonlinear compensation in the controller. The drive torque of the motor which is driving front wheel is output of the controller. The roll angle of body of VSBR and the angle that front wheel rotated are inputs of the controller. A multiple-input single-output (MISO) controller is built. The L2-gain of the system would be adjusted to make the controller robust. The simulation results of the RBF robust controller and the dynamic model of VSBR show feasibility and effectiveness of the dynamic model and the designed controller.