Mobile robot floor classification using motor current and accelerometer measurements

Abstract

Accurate localisation of an indoor robot critically depends on the odometry calibration which varies with different types of floor surfaces. Motion control accuracy of robots can be improved by independently calibrating odometry parameters for each floor surface. This paper presents a new robot floor classification system based on motor current measurements with compensation for variations in the floor inclination angles. The motor current is proportional to the rolling resistance on a flat floor when the robot travels at a constant velocity. We show that commonly occurring small deviations of less than one degree in the inclination of indoor floors significantly affects motor current measurements. The paper compensates for floor inclination variations with a low cost accelerometer. Floors are classified using a Support Vector Machine (SVM) with an accuracy of 95% for a 0.2 m travelling distance and 4 indoor surfaces that include similar carpets. Experimental results show that our proposed method significantly improves a previous floor classification system based on a colour sensor. Our previous work has shown that correct floor classification can improve robot motion control through more accurate odometry calibration, localisation, mapping and path planning.