Event Location: NSH 1507
Bio: Alberto Candela Garza is an M.S. in Robotics student at Carnegie Mellon University, advised by Prof. David Wettergreen. Alberto is affiliated to the Field Robotics Center and is interested in science autonomy for planetary rovers. Prior to CMU, Alberto received a B.S. in Mechatronics Engineering and a B.S. in Industrial Engineering from Mexico Autonomous Institute of Technology.

Abstract: As exploration goes to further extremes, communication becomes more constrained. Planetary rover operations are limited in bandwidth, delayed due to distance and resource restricted to just a few communication cycles per day. Despite being one of the closest planets from Earth, Mars rovers spend the vast majority of their time isolated, awaiting instructions. Clearly the exploration of much farther celestial bodies, such as the moons of Jupiter or Saturn, will be even more difficult. At the same time, planetary exploration involves frequent scientific reformulation and replanning. This problem suggests the need for a new paradigm where rovers can communicate and operate more efficiently by having a deeper understanding of the evolving scientific goals and hypotheses guiding their missions, rather than just collecting and sending data to human scientists for interpretation and planning.

This work establishes the science hypothesis map as a spatial probabilistic structure in which scientists initially describe their abstract beliefs and hypotheses, and in which the state of this belief evolves as the robot makes raw measurements. It discusses how to incorporate path planning for maximizing scientific information gain, which is efficiently computed. As proof of concept, this thesis describes a geologic exploration problem where a robot uses a spectrometer to infer the geologic composition of regions. This research shows that the science hypothesis map can be used to infer geologic units with high accuracy, and that exploration using information gain-based path planning has better performance than exploration with conventional science-blind algorithms.