Low level sensor fusion for autonomous mobile robot navigation

In this paper the development of a low-cost high precision navigation system for the autonomous mobile robot B21 is reported. True accelerometers providing not predictable time varying off-set errors have been additionally mounted on the robot to cope with the wheel skid and the wheel tracking error problems. The motor encoders of the B21 are reliable for indoor operation only if the former problems are not present. Therefore both sensor data have to be fused to obtain a satisfactory position estimate. This sensor fusion is implemented in the designed navigation system which combines low level sensor fusion with known statistics and rule-based sensor fusion. An improvement of the short and long term position estimate has been achieved.