Study on the Real-time RCT Neural Network Control Structure Using a Radial Basis Function Network for the Disturbance Rejection of a Gimbal System

In this paper, the roll angle balance control in the bicycle mode of a two-wheeled transformation robot with arms (TWT-A) for logistics is presented. TWT-A has two modes : Segway mode and bicycle mode. While in Segway mode, the pitch angle is controlled by two wheels, the roll angle in the bicycle mode is controlled by two synchronized reaction wheels. Since the reaction wheels are slow to respond to maintain balance, a disturbance observer (DOB) is designed to deal with any disturbing input to the roll angle balance. Balancing performance of DOB is evaluated through impact experiments. The disturbance observer reduced the disturbance by 47.7% more compared to the PID controller and showed the effect of suppressing vibration.