Abstract
This thesis develops an adaptive controller for a bicycle-riding robot that can provide consistent balancing and steering performance under uncertain center of gravity. The adaptive controller contains two parts: the adaptation part and the PD control part. While the adaptation part estimates the roll angle associated with the uncertain center of gravity, the PD control part uses the estimated roll angle and the measured roll rate to stabilize the bicycle laterally. The design of the adaptive controller is based on a linearized third-order model which contains the roll angle, the roll rate and the steering angle as the states. The input of the model is a fictitious quantity that once it is determined by the controller, the actual steering angle input to the bicycle is obtained by low-pass filtering the fictitious input. Simulations and experiments verify that the adaptive control can automatically steer the bicycle in a straight line even if mass imbalance exists.