AI Magazine - Winter 2014

A robot could even divert its path slightly to pursue science targets of opportunity (Woods et al. 2009).

Multiday plans could therefore make very efficient use of communications and personnel resources, enhancing long-distance survey missions.

The Life in the Atacama project is a NASA-spon-
sored effort to evaluate these techniques in the con-
text of desert subsurface biogeology (Cabrol et al.
2007). It uses Zoë (Wettergreen et al. 2008), a rover
capable of traveling more than 10 kilometers per day
and autonomously drilling up to 0.7 meter depth
(figure 1). As a mobility platform it combines navi-
gational autonomy with a changing payload of on-
board science instruments. Previous investigations
have used a fluorescence imager capable of detecting
very specific organic compounds and neutron detec-
tors to measure hydrogen abundance. The current
configuration incorporates a Raman spectrometer, a
visible near infrared point spectrometer, and naviga-
tion and science cameras. During a series of experi-
ments in the summer of 2013, scientists guided Zoë
remotely through the desert while exploring its geol-
ogy and biology.

This article describes the science autonomy system developed and tested with Zoë. It performed automatic acquisition of visible/near infrared (Vis-NIR) reflectance spectroscopy throughout the 2013 field season. This involved a range of different autonomous decisions exercised at various spatiotemporal scales. We begin by describing the rover platform and instrument payload. We then discuss instrument self-calibration, science feature detection, and targeting capabilities. We describe larger-scale path planning used to select informative paths between waypoints. We also detail the operational protocols used to command the rover and the results of its autonomous data collection. These experiments provide a case study of science autonomy deployed continuously over long distances. We report on system performance, lessons learned, and plans for future development.