Abstract: Uncertainties in environment and robot itself worsen performance of trajectory tracking control of a wheeled mobile robot. States of the robot cannot be correctly estimated by the odometry in this situation. The velocities estimation problem is described for a wheeled mobile robot without velocity sensors. Real-time velocities are estimated using accelerometers and position-measurement sensors. First the odometry signal is processed using sliding mode observer, then bandpass filter is applied to extract disturbances information of the robot body. The fused data in frequency domain is used to represent velocities of the wheeled mobile robot and fed into controller of an omnidirectional mobile robot with orthogonal-wheel assemblies for studying its trajectory tracking control. The experimental results show that more proper data of the robot states are provided and better control performance can be obtained using fused data.

Key words: Estimation, Mobile robot, Tracking control