New Position Estimation Method Based on Map Matching for Planetary Rover

This paper proposes a novel algorithm to precisely calculate the relative movement onboard of an autonomous rover based on 3D perception. It is a map matching technique that registers the visual information of consecutive rover positions and expresses the rover's displacement in a common coordinate system. Local elevation maps of the environment are therefore retrieved from the 3D perception of a stereo camera. The overall process breaks down into three main steps: feature point extraction from map data, matching of features points using triangle configurations and a voting procedure, and direct calculation of the rover displacement in six degrees of freedom using quaternions. The effectiveness of the proposed method was confirmed by computer simulations with synthetic terrain as well as experiments with stereo perception. Primary advantage of this approach in large unstructured terrain is that the utilization of many sensitive instruments can be avoided and large map displacements can be processed without any pre-registration.