Autonomous navigation of planetary rovers
- Autores: Martin Azkarate Vecilla
- Directores de la Tesis: Carlos Pérez del Pulgar (dir. tes.)
- Lectura: En la Universidad de Málaga ( España ) en 2021
- Idioma: inglés
- Tribunal Calificador de la Tesis: Aníbal Ollero Baturone (presid.), Anthony Mandow (secret.), Simon Lacroix (voc.)
- Programa de doctorado: Programa de Doctorado en Ingeniería Mecatrónica por la Universidad de Málaga
- Materias:
- Enlaces
- Resumen
From the experience and lessons learnt out of planetary rover missions such as the Mars Exploration Rovers or Curiosity, and looking forward to the upcoming missions of ExoMars and the Mars Sample Return programme, this thesis aims to study and enhance the autonomous navigation capabilities onboard the rovers, in order to increase the overall mission scientific return. As we see more and more autonomy being embarked in space missions, planetary rovers are already relying on self-navigating functionalities to fulfil their mission objectives. The constraints given by the space environment and the limitations in communications found in Mars exploration missions render the implementation of autonomous navigation capabilities as the most efficient solution, or the only, to extend the traversed distance per sol (a Martian day) beyond the few tens of meters.
Autonomous Navigation is a complex capability that relies on the implementation of several functionalities and their orchestrated execution in order to perform. Those functionalities are, as a minimum, Localisation, Perception or Mapping, Path Planning, and Path or Trajectory Control. In this thesis, these core robotics functions are investigated and newly developed methods that focus on the planetary rovers scenario are proposed. These methods take into consideration the constraints and conditions found in Mars missions, making them particularly fit and targeted for rover systems.
Different ways to combine and execute these functionalities are studied in order to compose the Guidance, Navigation and Control (GNC) subsystem of a rover that is capable of navigating autonomously. The potential variation in the terrain conditions found across a rover mission is taken into consideration, adapting the behaviour and functionalities run by the control architecture, with the objective of maximising the length of traverse per sol. A navigation mode is proposed targeting the relatively benign terrain cases. This mode relies solely on a stereo camera that mimics the ExoMars Localisation Cameras (LocCam) sensor to both localise the rover and avoid the hazards along the path. This is done without requiring periodic stops of the rover, increasing the effective traverse speed. The navigation mode is experimentally demonstrated in an exhaustive field test campaign totalling over 500m of autonomous traverse.
For rough terrain cases, a navigation mode that builds upon the first mode is proposed. It integrates a newly developed Simultaneous Localisation and Mapping (SLAM) component for planetary rovers that allows for improved accuracy of rover localisation and smoother detection and avoidance of terrain hazards. The map produced by the SLAM function is used in a Global Localisation component to correct the accumulated drift in long-range traverses. These two components are experimentally validated making use of a representative dataset gathered in a field test campaign run in a planetary analogue terrain near the Teide Volcano in the Canary Islands. Finally, the thesis proposes the design of a complete GNC architecture, that integrates the two navigation modes and could see a potential exploitation in future Mars missions such as the Sample Fetch Rover.
