Terrain traversability analysis using multi-sensor data correlation by a mobile robot

Abstract

A key feature for an autonomous mobile robot navigating in off-road unknown areas is environment sensing. Extraction of meaningful information from sensor data allows a good characterization of the near to far terrains, and thus, the ability for the vehicle to achieve its tasks with easiness. We present an image feature extraction scheme to predict mobile platform motion information. For a sequence of run, several images of terrains and vibrations endured by the mobile robot are acquired using a camera and an acceleration sensor. Texture information extracted by the Segmentation-based Fractal Texture Analysis descriptor (SFTA) was used to find correlations with acceleration features quantified using different time analysis parameters. Experimental results showed that texture information is a good candidate to predict running information.