Terrain coverage of an unknown room by an autonomous mobile robot

Abstract

Terrain coverage problems are nearly as old as mankind: they were necessary early in our history for basic activities such as finding food and other necessities. As our societies and their associated machineries have grown more complex, we have not outgrown the need for this primitive skill. It is still used on a small scale for cleaning tasks and on a large scale for {open_quotes}search and report{close_quotes} missions of various kinds. The motivation for automating this process may not lie in the novelty of anything we might gain as an end product, but in freedom from something which we as humans find tedious, time-consuming and sometimes dangerous. Here we consider autonomous coverage of a terrain, typically indoor rooms, by a mobile robot that has no a priori model of the terrain. In evaluating its surroundings, the robot employs only inexpensive and commercially available ultrasonic and infrared sensors. The proposed solution is a basic step - a proof of principle - that can contribute to robots capable of autonomously performing tasks such as vacuum cleaning, mopping, radiation scanning, etc. The area of automatic terrain coverage and the closely related problem of terrain model acquisition have been studied both analytically and experimentally. Compared to the existing works, the following are three major distinguishing aspects of our study: (1) the theory is actually applied to an existing robot, (2) the robot has no a priori knowledge of the terrain, and (3) the robot can be realized relatively inexpensively.