Probabilistic Method for Mapping & 3D SLAM of an off-road Terrain with Four Wheeled Robot

Abstract

Mobile robots use onboard range sensors and accurate, real-time mapping to perform autonomous navigation in challenging terrain. Absolute localisation based on the tracking of exterior geometric or visual cues is frequently used in existing techniques. To overcome the dependability issues with current methods, we propose a novel method for terrain mapping that only uses interoceptive clustering from kinematic and inertial parameters. The suggested approach takes into account the state estimation's drift and uncertainty as well as noise model for the distance sensor. As probabilistic terrain estimate, it generates a grid-based altitude projection with both higher and lower confidence bounds. We demonstrate the effectiveness of our approach for legitimate terrain mapping with wheeled robots using simulated datasets and practical experiments, and we compare the terrain rebuilding with referenced maps that depict the ground reality.