Towards learning-based trajectory tracking control for a planetary exploration rover: Development and testing

Abstract

Slip estimation and compensation are crucial factors for the success of ground-based planetary exploration missions. This work presents the development and implementation of a test campaign for the systematic evaluation of trajectory tracking controllers for a planetary exploration rover. The performed tests focus on machine learning algorithms for wheel-ground interaction. The control approaches are developed and tested both in simulation and on a rover prototype in a space-analog scenario. First results comparing the performance of the classical tracking controllers and their learning-based extensions are presented. The results imply that the use of machine learning techniques can improve the tracking performance on difficult terrain significantly.