Science Requirements for a Planetary Robotic On-Board Vision Processing System.

The robotic exploration of planetary surfaces, including missions of interest for geobiology (e.g., ExoMars), will be the precursor of human missions within the next few decades. Such exploration will require platforms which are much more self-reliant and capable of exploring long distances with limited ground support in order to advance planetary science objectives in a timely manner. The key to this objective is the development of planetary robotic on-board vision processing systems, which will enable an autonomous on-site selection of and access to scientific and mission-strategic targets. The EU-funded research project PRoViScout (Planetary Robotics Vision Scout) is designed to develop a unified and generic approach for robotic vision on-board processing, namely the combination of navigation and scientific target selection. Any such system needs to be "trained", i.e. it needs (a) scientific requirements which the system needs to address, and (b) a data base of scientifically representative target scenarios which can be analysed. We present our preliminary list of science requirements, based on previous experience from landed Mars missions, and we demonstrate selected examples of the training scenarios/images, built from.