Balance control for two-wheeled robot via neural-fuzzy technique

A neural-fuzzy-based balance controller for two-wheeled robot is proposed in this paper. In the fuzzy controller, the total sliding surface is adopted as the input variable of fuzzy system to outstanding the merit of its insensitivity to uncertainties. In the fuzzy membership function, the translation width idea is utilized to reduce the chattering phenomena. Moreover, consider the parametric variation, external disturbance and nonlinear friction for the practical wheeled robot motions, the transient and unmodelled uncertainty will be occurred. So, a hetero-associative neural network, which is utilized to observe the uncertainty, is added into the controller to reduce the accumulated error and to ascend the stability. The hardware includes a microcontroller, gyroscope, accelerometer, and two autonomous motors, etc. The effectiveness is verified by experimental results, and the performance is compared with conventional PD control schemes for the same wheeled robot.