Information-Driven and Risk-Bounded Autonomy for Scientist Avatars

Abstract

We present an overview of an information-seeking risk-bounded planning and execution system to enable future spacecraft to make decisions and adapt their behavior to seek out the most high-value scientific information, while bounding risk of failure. Information-seeking autonomy is an innovation with potential for transformational impact on the way our spacecraft enable scientific discoveries, by complementing traditional scientist-in-the-loop operations with appropriately-conservative onboard direction of science measurement activities (based on scientist-specified models). Our executive selects a measurement strategy that maximizes information based on the environment and initial measurements. It takes various mission risks into account when planning and executing activities to deal with uncertainties and disturbances. It can adapt its strategy based on collected measurements onthe-fly. Finally, it demonstrates resilience, despite failures and degradations, to achieve its high-level scientific objectives. We have performed an initial demonstration of the capabilities of our executive against a basic spacecraft simulation of an asteroid flyby scenario, leveraging the Robot Operating System and the Basilisk open-source simulation framework. It has also been demonstrated on autonomous underwater vehicle scenarios.