Engineering topics
Wagstaff, Kiri
Publications and source records attributed to Wagstaff, Kiri.
Benchmarking Machine Learning on the Myriad X Processor Onboard the ISS
No abstract provided
Mars Image Content Classification: Three Years of NASA Deployment and Recent Advances
The NASA Planetary Data System hosts millions of images acquired from the planet Mars. To help users quickly find images of interest, we have developed and deployed contentbased classification and search capabilities for Mars orbital and surface images. The deployed systems are publicly accessible using the PDS Image Atlas. We describe the process of training, evaluating, calibrating, and deploying updates to two CNN classifiers for images collected by Mars missions. We also report on three years of deployment including usage statistics, lessons learned, and plans for the future.
Content-based On-board Summarization to Monitor Infrequent Change
No abstract provided
Time-Series Methods for Responsive Instruments: A Flight-System-Like Implementation
No abstract provided
Detection of Environment Transitions in Time Series Data for Responsive Science
No abstract provided
Efficient Active Learning for New Domains
The promise of active learning is to reduce the number of labeled examples required by supervised machine learning algorithms. The largest potential benefits lie in entirely new domains, for which no labeled examples yet exist. Yet to date, most active learning studies are retroactive and demonstrate the benefits that could have been gained if active learning had been used. What are the barriers to true adoption and utilization of active learning? We focus on two: (1) the cold start or class discovery problem, in which active learning methods may struggle to make progress with zero labeled examples, and (2) the cost of having the classifier in the loop to select the next example to be labeled. We assess different active learning approaches in the context of these two barriers and conclude with recommendations for how to employ active learning in new domains. As an example, we report on the use of active learning on a large, novel data set of Mars surface images.
ARIEL: Autonomous Excavation Site Selection for Europa Lander Mission Concept
This paper presents ARIEL (Autonomous Ranking and Interrogation of Excavation Location), an autonomy system for selecting an excavation site on-board for NASA’s Europa Lander Mission Concept. Historically, excavation site selection has been performed by a lengthy ground-in-the-loop (GITL) process involving manual inspections, assessments, and decision making in past missions. However, as Europa Lander would have approximately 20 days of lifetime after the landing, many surface activities, including excavation site selection, must be autonomously performed on-board. This paper describes the overall system of ARIEL as well as its two major algorithmic components: vision-based candidate selection and smart interrogation, which estimates the physical properties of the icy surface through physical contact with the robotic arm’s endeffector. Preliminary results are presented using images from Earth analogue sites. The Europa Lander mission returned to the formulation phases in early 2019 while ARIEL was at an early stage of development. Described in this paper is a snapshot of ARIEL as of the project suspension. This paper also describes the remaining challenges to be solved, should the mission resume in the future.
Scientist-Guided Autonomy for Self-Reliant Rovers
Although Mars rover missions have been highly successful in accomplishing scientific objectives, mission productivity is limited by challenges stemming from the need for commanding ground-based targeted observations under communication constraints imposed by the large distance between Earth and Mars. With an aging fleet of sun-synchronous relay orbiters, the opportunities for regular communication with rovers may become even more limited. In addition to on-board planning, robust navigation, and health assessment, there are strategies to make future rovers more self-reliant by enabling them to perform autonomous scientific characterizations of new areas during periods without an opportunity for ground-based targeted observations. In particular, we have studied how a ”walkabout” strategy, in which an initial high-level characterization of a region is used to informed subsequent passes with specific targeted observations, was used successfully during the investigation of Pahrump Hills by the Mars Science Laboratory. Inspired by this approach, we have identified several capabilities that could allow a rover to autonomously perform some of these initial high-level characterization steps. In this paper, we describe technologies for identifying specific geologic units, regions, or features of interest, identifying areas of contact between two adjacent units, detecting and determining the orientation of layering within rock units, identifying novel and interesting features, and planning observations of regions with different sampling strategies using remote sensing instruments. The observations acquired with these approaches are driven by scientists’ guidance and can provide scientists with data to help inform their decisions about where to make more resource-intensive targeted observations.
A Case Study of Productivity Challenges in Mars Science Laboratory Operations
Achieving consistently high levels of productivity has been a challenge for Mars surface missions. While the rovers have made major discoveries and dramatically increased our understanding of Mars, they often require a great deal of effort from the operations teams and achieving mission objectives can take longer than anticipated. We conducted an in-depth case study of Mars Science Laboratory operations in order to identify the productivity challenges facing surface missions. In this paper, we describe how we performed the case study and analyzed the data. We present and discuss the significant productivity challenges we identified during the study. In addition to informing future surface exploration missions, the study is relevant for a wide range of applications in which operators must interact with a robotic system with limited communication opportunities.
Flight validation of instrument processing onboard Earth Observing One: a status report
UNKNOWN