Some sensing and perception techniques for an omnidirectional ground vehicle with a laser scanner

This paper presents some techniques for sensing and perception for an omnidirectional autonomous ground vehicle equipped with a laser scanner. In an assumed structured environment, the sensor data processing methods for both 1D and 2D laser scanners, are discussed. Raw data are segmented to lines, circles, ellipses, planes and corners by task dependent segmentation algorithms. Each subset of data is then used for fitting the known template shapes as listed above. An arbitrator is used to discriminate the shape for the final object identification.