Vision-Based Rover Navigation Demonstrator

Abstract: This work is part of the preliminary stages of the PRoViScout project. PRoViScout is a collaboration of the major European groups in the field of robotic exploration, with the aim to further the autonomous capabilities of planetary exploration rovers. The following Master Thesis report describes the creation, functionality and results of a vision-based rover navigation demonstrator, designed to practically test the current sensors onboard the PRoViScout rover in a navigation scenario. The work has a focus on implementing the novel 3D time of flight sensor into the vision system but also describes the continued integration of a stereo camera pair, based on earlier work done by the hosting institute Joanneum Research. For comparison the report gives a description of the vision and navigation systems of the only practically implemented and successful planetary rovers; Sojourner, Spirit and Opportunity. A collective description is also given on the experiences gained from various projects implementing a 3D time of flight sensor.

The methods and tools developed by the host institution are described, as well as the newly developed methods during this work, primarily aimed toward the implementation of a 3D time of flight sensor and navigation functionality. The components developed in this work consist mainly of sensor interfaces, mapping of sensor data into digital elevation maps and a navigation application making use of these tools. The results from the components are shown and analyzed to derive various strengths and weaknesses. A discussion on the characteristics of the system is given as a basis for a conclusion.

The results show a functional prototype, suitable for the purpose of acting as a vision-based rover navigation demonstrator. Some of the improvements required to advance to the stage of a robust end-system, usable for planetary missions, are noted.