Design and development of a stereo vision-based navigation and guidance system for a space rover

The lag in telecommunications between a planetary exploration rover and the Earth does not allow the real-time control of the platform. Autonomous guidance and navigation capabilities are therefore mandatory requirements for planetary exploration missions. This paper deals with the definition, design, assembling and testing of a low-cost guidance and navigation system intended to be into the rover RAGNO (Rover for Autonomous Ground Navigation and Observation), an autonomous platform developed at Sapienza Università di Roma to familiarize with vehicles operating in hazardous or poorly known environment. The navigation function, while possibly completed by magnetometers and additional sensors, is mainly based on visual techniques, as they can reliably offer the required accuracy with suitable cost and complexity. For the purpose of obstacles identification and localization, basic to rovers operating in unknown environment, a stereoscopic navigation system has been selected. The observation from two points of view allows to localize a detected object by means of a simple triangulation of corresponding point pairs (features). The basic relations and the techniques implemented to detect and to match the features of the two images are briefly explained. As a result, the rover is able to autonomously understand the surrounding scenario. Once the obstacles have been localized, a safe and optimal path to reach the desired targets must be planned. To this aim, the implementation of a guidance algorithm based on manifold cell discretization, graph theory and length optimization has been presented. This algorithm, a customization of the A-star search method, is compared to a more classical one, based on the Lyapunov approach with the introduction of artificial potential functions centered in the identified obstacles.