Opportunistic Detection and Measurement of Novel Rocks

Introduction: The Onboard Autonomous Science Investigation System (OASIS) evaluates geologic data gathered by a planetary rover [1]. This analysis is used to identify new science opportunities during a rover traverse. A planning and scheduling component enables new science measurements to be safely added to the rover current command sequence. Recent tests have highlighted the identification of novel terrain features during a rover traverse. This capability has been recently added to the OASIS system and has been successfully demonstrated using the FIDO rover at NASA's Jet Propulsion laboratory. In addition, the system has been integrated with the MER rover mission Visual Target Tracking (VTT) capability, which enables the system to robustly track a specified target while the rover drives. By integrating this capability, the rover can be commanded to approach targets identified onboard and acquire targeted measurements both from additional viewing angles as well as from positions in close proximity to the target. Novelty Detection: During a rover traverse, OASIS periodically acquires Navcam and Hazcam images. When an image is acquired, it is analyzed to identify rocks in the image and, for each rock, geologically salient features including albedo, size, eccentricity and angularity (shape), and orientation are determined. Each rock and its features are added to an onboard database that records the history of the traverse. Using this database, the OASIS novelty detector identifies rocks that are dissimilar from any rocks seen previously [2]. The novelty detector i) groups rocks based on feature similarity, and then ii) searches for statistical outliers from those groups. Intuitively, this has the effect of identifying and prioritizing for downlink the most novel rocks. For example, a novelty detector that utilized albedo or VIS/NIR spectral signature would identify a calcite as novel in a sea of martian basalts. The novelty detector groups rocks based on feature similarity. To perform this grouping, rock features are treated as points in N-dimensional space and then clustered using the unsupervised k-means clustering algorithm. Once the N