Sensory navigation of autonomous cleaning robots

This paper addresses a general problem of sensory navigation for autonomous cleaning robots (ACR) in unknown environments. Although absolute self-localization and accurate environment modeling still remain unsolved, a delicately designed mobile robot still can carry out a task practically. Under this strategy, we propose a cleaning robot system, which works without any environment map and global self-localization. It has a three-layer structure. The lower layer is composed of general hardware: ultrasonic sensors, infrared sensors, incremental encoders, DC motors, vacuum, etc. Upon these sensors is the sensory behavior layer, which includes several motion templates. These intelligent motion templates can deal with most situations, seldom making the robot trapped. The upper layer is the task-based navigation layer, which carries out the tasks of environment learning, cleaning, and homing. Finally the experiment results show that this strategy works well, even if the robot knows only a little about the environment.