Description: Robotic explorers today are primarily controlled by detailed plans exhaustively created by a large group of human scientists and engineers. For example, the Mars Science Laboratory rover is operated by teams of people that re-plan the rover actions each Martian sol. As robotic explorers are sent further into the Solar System, limited bandwidth, communication delay, and the number of active missions make human-centered planning increasingly impractical and, in some cases, infeasible. A rover exploring the subsurface oceans of Europa will likely have low bandwidth overall with periods of communication blackout, and a short operational life to accomplish its mission. Therefore, it is imperative to employ robot autonomy systemically to improve productivity in all aspects of the life of a mission. Autonomy on a system-wide scale that reflects how human scientists behave in the field would significantly increase the scientific information gained from a mission. Human scientists continually reinterpret their measurements and assumptions with a growing contextual knowledge of the environment. They use each new piece of data to inform the next action while cognizant of operational constraints such as time, energy, and risk. State-of-the-art autonomous exploration methods expect a consistent objective and do not update the plan with each new measurement and changing priorities. The idea of systemic autonomy is extending beyond instrument-level autonomous functions to system-level autonomy, which enables the robot to reason about the environment, interpret science goals and hypotheses, determine utilization of all its instruments, monitor health and resources, and recover from faults. These functions allow the robot to make decisions regarding sampling and measurement to maximize information gained. Systemic autonomy enables exploration to proceed without human assistance as well as enhances robotic explorers as tools for human scientists. We will research and develop systemic autonomy for robotic explorers to increase information gain while respecting operational constraints. We will focus on scientific exploration with robots acting as proxy to realize the high-level science objectives of human investigators. We will leverage prior work in robotic planning, including factorized particle filters, the Rao-Blackwell representation, and information theoretic criteria for path planning and action selection. We will harness recent work on Science Hypothesis Maps that extend a classical mapping framework to spatially represent the scientist’s interpretation and objectives quantitatively. The unique contribution of our work will be to incorporate scientific belief state into robotic decision making, including action planning. The robot will be able to automatically analyze new data to update this informative representation, plan trajectories, and plan scientific measurements in a resource efficient manner. The NASA Technology Roadmap expresses the need for better onboard robot autonomy. This was also endorsed at the 2018 Workshop on Autonomy for Future NASA Science Missions. This research addresses the Robotics and Autonomous Systems high priority topic of on-board planning and scheduling while managing resources (TA4.5.2). Automatically analyzing scientific measurements to inform sampling decisions and path trajectory is relevant to the topic of autonomous decision making (TA4.5.8) and intelligent data collection and prioritizing data based on content (TA11.4.2). This research is also relevant to the topic of adaptive systems frameworks (TA11.4.5) that can manage a set of interacting dynamic goals and constraints to continually update trajectories and measurement decision

Benefits: The NASA Technology Roadmap expresses the need for better onboard robot autonomy. This was also endorsed at the 2018 Workshop on Autonomy for Future NASA Science Missions. This research addresses the Robotics and Autonomous Systems high priority topic of on-board planning and scheduling while managing resources (TA4.5.2). Automatically analyzing scientific measurements to inform sampling decisions and path trajectory is relevant to the topic of autonomous decision making (TA4.5.8) and intelligent data collection and prioritizing data based on content (TA11.4.2). This research is also relevant to the topic of adaptive systems frameworks (TA11.4.5) that can manage a set of interacting dynamic goals and constraints to continually update trajectories and measurement decision