Rover relative localization testing in martian relevant environment

In this work we describe a relative localization algorithm developed to estimate the position and the attitude (roll, pitch and heading angle) of a rover on the surface of Mars. The algorithm is designed to operate on a ground control center and to exploit all the computational power which is not available on the on-board computer and it serves to refine the rover's telemetry data. The input consists of a set of stereo-images captured by the rover during operations and separated by at least 1 m apart from each other; limitations in the number of downlinked images, due to the variable and reduced bandwidth, has been a driver for the design.The performances of the algorithm in term of accuracy have been tested by capturing a series of stereo-images in a representative martian environment, the ALTEC Mars Terrain Simulator, with a half-scale version of the ExoMars rover, called ExoMars Testing Rover (ExoTeR, property of ESA, European Space Agency). The ground truth was given by a motion capture system. We demonstrate the algorithm capability to estimate the rover pose even in case of significant camera shifts.