An Approach to Autonomous Science by Modeling Geological Knowledge in a Bayesian Framework

摘要

Autonomous Science is a field of study which aims to extend the autonomy ofexploration robots from low level functionality, such as on-board perceptionand obstacle avoidance, to science autonomy, which allows scientists to specifymissions at task level. This will enable more remote and extreme environmentssuch as deep ocean and other planets to be studied, leading to significantscience discoveries. This paper presents an approach to extend the high levelautonomy of robots by enabling them to model and reason about scientificknowledge on-board. We achieve this by using Bayesian networks to encodescientific knowledge and adapting Monte Carlo Tree Search techniques to reasonabout the network and plan informative sensing actions. The resulting knowledgerepresentation and reasoning framework is anytime, handles large state spacesand robust to uncertainty making it highly applicable to field robotics. Weapply the approach to a Mars exploration mission in which the robot is requiredto plan paths and decide when to use its sensing modalities to study ascientific latent variable of interest. Extensive simulation results show thatour approach has significant performance benefits over alternative methods. Wealso demonstrate the practicality of our approach in an analog Martianenvironment where our experimental rover, Continuum, plans and executes ascience mission autonomously.