Next-generation planetary exploration rovers will soon be able to travel long distances in a single-communication cycle. This offers an enormous opportunity but also an important challenge: robots will need to do science autonomously and understand collected data in order to make more effective exploration decisions. We have developed a system to obtain spectroscopic measurements on-the-fly, pointing autonomously at geologic targets from the distance using a reflectance spectrometer. Our technology allows a robot to detect surrounding rocks, potentially get hundreds of precise spectral measurements of targets and validate the data, all in one command cycle with no direct human intervention. We have fully implemented our system on Zo?robot, where we have proved to successfully track and aim a variety of geologic targets. In this work we describe the method, calibration and procedure for spectrometer pointing, explain our approach detecting and localizing rocks and show how we classify targets as rocks or soils based on spectrometer data. These results are relevant for science autonomy systems to help in the buildup of geologic maps of the traversed regions.

Science on the Fly Project