Abstract
Current techniques for the exploration of planetary surfaces are slow, and involve frequent human intervention. Rapid growth in complexity of future missions requires the development of ambitious technologies that may allow remotely deployable vehicles to carry out a majority of their tasks autonomously. Mapping and localisation is a key area of planetary exploration that can benefit from an increased level of autonomy, for instance in upcoming missions such as ExoMars and the ESA sample return proposal FASTER. Simultaneous Localisation And Mapping (SLAM) is a technique developed for terrestrial uses allow autonomous vehicles to calculate their position in an unknown environment, whilst also creating a map. This technique can be transferred to a planetary platform in order to allow the rover to better traverse through its environment without human intervention. While several techniques existed for implementing SLAM in a terrestrial environment, the use of SLAM for planetary exploration has not been explored in depth. The use of LIDAR, commonly used in terrestrial SLAM implementations, is a complex solution for use in space missions, and has yet to be robustly proven. Therefore, another means of observing the environment, a key part of the SLAM algorithm, is required. Monocular cameras together with vision processing algorithms present a simple scenario well suited to the scenario. The work presented in this paper comprises the design and implementation of a fully end-to-end, modular planetary SLAM system. The system takes data from the Planetary and Asteroid Natural scene Generation Utility, and using vision based algorithms passes observation data to one of three SLAM filters, the Extended Kalman Filter, the Extended Information Filter and FastSLAM. The results show that these techniques and filters are well suited to the planetary environment and provide a route towards extending rover autonomy. ©2013 by the International Astronautical Federation. All rights reserved.