Estimating a room size using encoders and collision detectors: application to a cleaning mobile robot

Floor cleaning is a typical mobile robot application. Most commercial cleaning mobile robots designed for domestic applications use random path-planning algorithms and very few sensors and the critical problem is to decide when the cleaning is already finished or the coverage completed. In this paper, we propose some methods to estimate the size of the cleaning area using only the information of encoders and collision detectors. To this end, three typical rooms have been tested with and without furniture. Results show the great influence of furniture distribution in the estimation of the area of the room. Average estimation errors under plusmn20% can be expected for empty rooms and from 0% to -70% in the case of rooms with furniture.

I. Introduction

Floor cleaning is a noncreative human task that can be automated using mobile robots. Coverage algorithms based on genetic algorithms [3], [4], neural networks [5], [6], exact cell decomposition [15], spanning trees [7], spiral filling paths [8], [9], [10], etc, are highly sensitive to the internal [16], [17] and external [21], [22], [23] sensors of the mobile robot and the previous or acquired information of the area where the robot operates [1], [2]. Moreover, the typical domestic cleaning scenario becomes unstructured and unknown by the furniture, decoration things and the typical human disorder habits. Then the complete-coverage problem becomes more and more complex [11]–[15] and high quality sensors as ultrasonic or laser are needed to obtain a map [18]–[20] of the cleaning scenario.