Abstract
Future planetary exploration missions on the surface of distant bodies such as Europa or Enceladus can’t rely on human-in-the-loop operations due to time delays, dynamic environments, limited mission lifetimes, as well as the many unknown unknowns inherent in the exploration of such environments. Thus our robotic explorers must be capable of autonomous operations to ensure continued operations and to try to maximize the amount and quality of the scientific data gathered from each mission. To advance our technology toward this goal, we are developing a system to maximize the science obtained by a robotic lander and delivered to scientists on Earth with minimal asynchronous human interaction. The autonomy architecture consists of three main components- Shared Science Value Maps (SSVMs), which function as an interface between REASON (Robust Exploration with Autonomous Science on-board) and RECOURSE (Ranked Evaluation of Contingent Opportunities for Uninterrupted Remote Science Exploration) for efficient and useful scientific communication between scientists and robot. The key advantage to this design is in its ability to continuously operate and adapt despite the constraints of high-latency, low-bandwith communications and an uncertain environment which today would require ground-in-the-loop operations. This paper presents the overview of our architecture and initial results on the development of such a system. These results will focus on progress made in developing the details of the SSVM interface between human scientists and robotic explorer and the ability of REASON to act on the SSVM to develop plans on-board that attempt to maximize science obtained while being guaranteed to respect any relevant system and safety constraints.
Karan Muvvala
Graduate Research Assistant
Building safer and smarter robots using Game-theoretic approaches while leveraging techniques developed by the Formal methods community to provide safety-critical guarantees.