Localization of 3D objects based on ultrasonic data, considering the influence of measurement uncertainty

Abstract

Map building is still an important preparation for mobile robots to navigate in initially unknown environments. Primarily low costs are responsible for the use of ultrasonic sensors for this purpose. Since many years work is invested in this field but most approaches are 2D, assuming orthogonal objects and a planar robot movement. In this paper a more general method to localize objects in arbitrary pose is introduced. The well-known problem of unknown echo direction inside wide sensor lobes is answered by using the lobe boundaries. Specular reflections, which are problems in usual, are required attributes of objects for the pose calculation. Measurement uncertainties in the lobe boundary and distance measurements have great influence on this method. Thus the error sources and their effects are shown. Improvements in sensor systems however are promising to overcome the weaknesses of industrial sensors used in this work.