Abstract
This paper presents landmark based self-localization of a two-wheeled differential drive autonomous mobile robot in a known but highly dynamic environment. The robot is equipped with a pivoted stereo vision system, two digital encoders, a gyro sensor, two 10g accelerometers and a magnetic compass. Global position of the robot is estimated using range measurements of distinct features such as color transitions, corners, junctions and line intersections in the robot environment. However, due to scarcity of distinct features, it is not possible to extract the minimum required features for global position estimation from everywhere in the state space. Therefore, the robot position is tracked between intermittent global localization to have an all time position estimate available to the robot. The robot observation vector is composed of range and bearing measurements of distinct features in the robot environment which is merged with the current position estimate to suppress the accumulating errors. Simulation results illustrate the performance of the location tracker.