Learning for mobile robots: environment map acquisition

The mapping of environment using sensory information has generated a great deal of interest. This paper presents a technique for learning, which allows the mobile robot to acquire the environment map automatically. The technique is based on the principle of explicitly representing and reasoning about the uncertainty in robot´s position and orientation, as well as uncertainty in sensing process. The learning procedure is composed of matching and merging. Matching is to compare the sensed-map with the known-map of the robot to find a set of correspondences; merging is to incorporate new or more accurate information in sensed-map into known-map using the correspondences for guidance. By performing matcher and merger constantly, the robot will ultimately acquire the complete map of its environment