Life in the Atacama: Science autonomy for improving data quality

Abstract

“Science autonomy” refers to exploration robotics technologies involving onboard science analysis of collected data. These techniques enable a rover to make adaptive decisions about which measurements to collect and transmit. Science autonomy can compensate for limited communications bandwidth by ensuring that planetary scientists receive those images and spectra that best meet mission goals. Here, we present the results of autonomous science experiments performed in the Atacama Desert of Chile during the Life in the Atacama (LITA) rover field campaign. We aim to provide an overview of autonomous science principles and examine their integration into the LITA operations strategy. We present experiments in four specific autonomous science domains: (1) autonomously responding to evidence of life with more detailed measurements; (2) rock detection for site profiling and selective data return; (3) tactical replanning to efficiently map the distribution of life; (4) detecting novel images and geologic unit boundaries in image sequences. In each of these domains we demonstrate improvements in the quality of returned data through autonomous analysis of imagery.

Application of pulsed‐excitation fluorescence imager for daylight detection of sparse life in tests in the Atacama Desert

“…These include a single FOV of either C‐FI or F‐FI, three side‐by‐side FOVs of F‐FI which could be mosaiced together (see Figure 6), a five‐position set of C‐FI or F‐FI FOVs for greater coverage of a locale, and C‐FI or F‐FI FOV sets of a plow trench, acquired before and after the plow operation and positioned in the center and on the margins of the trench. During the 2005 field season, Science on the Fly periodic sampling was introduced [ Smith et al , 2007]. With that procedure, every C‐FI of the transect up to the remote science team's defined limit was followed up with an F‐FI if on‐board image processing indicated a positive signal.…”

The Coevolution of Life and Environment on Mars: An Ecosystem Perspective on the Robotic Exploration of Biosignatures

“…Another means of supporting biosignature exploration prior to, and during, a mission is by integrating survey techniques developed in microbial ecology. Past and recent NASA-funded rover field experiments testing microbial habitat and biosignature exploration strategies at the surface and in the subsurface of the Atacama Desert ( e.g., Wettergreen et al, 2005 , Cabrol et al, 2007a ) led to the development of ecological mapping techniques adapted to autonomous rover operations, and to the generation of predictive maps of habitat potential ( e.g., Hock et al, 2007 ; Warren-Rhodes et al, 2007a , 2007b , and unpublished data; Smith et al, 2007 ). This is typically an area where advances in detection algorithms and machine learning can bring new support to biosignature detection, including from orbit through, for example, the segmentation of hyperspectral imagery, correlation of orbital and surface spectra, novelty detection, multiscale adaptive sampling, data segmentation and superpixel clustering, and biohint probability mapping, among others ( e.g., Thompson et al, 2011 ; Candela et al, 2017 ).…”

Robotic astrobiology – prospects for enhancing scientific productivity of mars rover missions

“…The Viterbi algorithm generated a maximum-likelihood image sequence that maximizes the posterior probabilities of scientific value (quantified as an information metric). The most comprehensive experiments in autonomous science (in astrobiology) have been conducted in the Atacama Desert using the Zoe rover test platform (Smith et al 2007). It used spectral signatures (such as chlorophyll) detected during traverse classified by a Bayesian net to trigger the rover to stop to collect further sample data.…”

“…The most comprehensive experiments in autonomous science (in astrobiology) have been conducted in the Atacama Desert using the Zoe rover test platform (Smith et al . 2007). It used spectral signatures (such as chlorophyll) detected during traverse classified by a Bayesian net to trigger the rover to stop to collect further sample data.…”