CLAWS · UNIVERSITY OF MICHIGAN × 2020 NASA SUITS CHALLENGE

An AR field kit for the Moon.

ATLAS — the AR Toolkit for Lunar Astronauts and Scientists — is a modular HoloLens system for Artemis-generation lunar EVAs, paired with VEGA, a Rasa-based conversational AI, and the CMCC mission-control web app. Designed by a 30+ member interdisciplinary student lab, tested with NASA scientists and flight controllers, and winner of the 2020 NASA SUITS Challenge.

ROLEHuman-in-the-Loop (HITL) Testing Coordinator · UX Research
TEAM30+ students · advised by Dr. Leia Stirling · UI/UX consults with Dr. Michael Nebeling
METHODSExpert interviews, task flowcharts, XD/Figma prototyping, NASA-TLX test design
OUTCOME2020 NASA SUITS winner · $10K Epic MegaGrant · foundation for all later CLAWS systems

The challenge

NASA's SUITS challenge asks university teams to design the visual display system for the xEMU — the Artemis-generation spacesuit — replacing the cuff checklists and wrist-mounted mirrors astronauts still use today. The brief is unforgiving: support navigation, sample collection, repairs, and emergencies on the lunar surface without inhibiting the mission.

A spacesuit display competes for attention an astronaut cannot spare. Every design decision started from that constraint: non-intrusive first, everything else second.

CLAWS first competed in 2019 with MEISSA, an ISS-focused predecessor tested with EVA flight controllers at Johnson Space Center. For 2020 the team grew from 10 to over 30 members, and the problem moved to the Moon: drastic lighting swings, unfamiliar terrain, and geology work that demands both hands.

ATLAS Visor UI — Interaction States Web recreation of the standard non-intrusive view
COLOR CODES PER NASA CONVENTION — AMBER / RED SEVERITY · PURPLE = STALE TELEMETRY

Grounding the design

Before a single mockup, the team ran a literature review of EVA workflows and modeled how astronauts navigate to a site and collect samples as task flowcharts. Then we went to the people who'd actually done the work — six expert engagements that directly shaped the system:

OCT 2019

Jacob Richardson — planetary volcanologist, NASA Goddard

Introduced geology field notes; recommended analog field-trek sites near Michigan.

NOV 2019

Brent Garry — planetary volcanologist, NASA Goddard (BASALT, DRATS)

Minimap navigation and comms architectures → shaped the navigation module and CMCC communications.

DEC 2019

Ben Feist — software engineer, "Apollo 17 in Real Time"

Field notebooks, data storage and syncing → foundational to the CMCC's database design.

JAN 2020

Trevor Graff — Chief Scientist, ARES group, JSC · NEEMO 22 aquanaut

AR in extreme analog environments; NASA sampling processes → greatly influenced the GeoNotes sampling module.

FEB 2020

Dr. Jessica Marquez — Human Systems Integration, NASA Ames

Notifications, messaging, and text-to-speech; granted research access to Playbook, integrated into the CMCC.

ONGOING

Geology graduate students

Interviews on sampling tasks, field-note conventions, and rock-description vocabulary.

Assessing perception and cognition requirements led to a hard rule: the interface could carry no ambiguity. Icons were barred from the final mockup and replaced with proper terminology.

Other decisions followed the same logic. The color scheme was engineered to stay visible across the drastic lighting swings of the lunar surface. Typography followed NASA's recommendation (Helvetica headers). And the warning hierarchy was matched to the patterns used by the Michigan Medicine emergency department, which has well-tested rules for when and how to interrupt busy people.

GeoNotes — Sample Collection Protocol From the ATLAS UX flowcharts
01

Navigate

Minimap + waypoints planned pre-EVA in the CMCC guide the astronaut to the site.

"VEGA, take me to waypoint 3"
02

Image the site

Photograph the surrounding environment before disturbance, per NASA sampling process.

03

Collect & store

Guided procedure for collecting the sample and stowing it with a sample identifier.

04

Record GeoNote

Digital field notes — location, timestamp, author, category — replace paper notebooks.

"Log: vesicular basalt, crater rim"
05

Sync to CMCC

Data streams to mission control for scientists to review live and in post-ops debrief.

The system

ATLAS is deliberately modular. Every capability is a module — the smallest unit of functionality. Modules compose into protocols, each catered to a phase of the EVA life cycle: mission planning, suit prep, sample collection, rover repair, emergency warnings, abort. A central Protocol Manager drives the application, swapping protocols while modules hold their state — so an astronaut moving from sample collection to an abort warning keeps the same biometrics and navigation in view, with no reload and no lost context.

01 Navigation02 Biometrics 03 Mission tasks04 Communications 05 GeoNotes06 Sample ID 07 QR scanner08 Warnings

Around the headset sits a full sociotechnical system: the CMCC — a Django/PostgreSQL mission-control web app, migrated to the cloud during COVID — supports planning (mission tasks, waypoint marking), live operations (control room, biometric monitoring, "get-ahead" tasks), and post-ops debrief (field notes, recordings, route maps). A Raspberry Pi sensor suite (GPS, IMU, photoresistor-triggered lighting) stands in for lunar positioning infrastructure.

System Architecture ATLAS · VEGA · CMCC · Sensor suite
ON THE ASTRONAUT — MICROSOFT HOLOLENS ATLAS AR Toolkit for Lunar Astronauts & Scientists PROTOCOL MANAGER PROTOCOL: SAMPLING PROTOCOL: REPAIR · ABORT… MODULES (PERSIST ACROSS PROTOCOLS) NAV · BIOMETRICS · TASKS · COMMS GEONOTES · SAMPLE ID · QR · WARNINGS VEGA Rasa conversational AI — voice in, voice out ON THE GROUND — CLOUD (HEROKU) CMCC CLAWS Mission Control Center — Django + PostgreSQL PRE-OPS Mission tasks · waypoint planning OPERATIONS Control room · biometrics · get-ahead tasks POST-OPS Field notes · recordings · route map IN THE EVA BACKPACK RASPBERRY PI SENSOR SUITE NEO-6M GPS · IMU · photoresistor light control JSON / HTTP — telemetry, GeoNotes, media waypoints · tasks · warnings · get-aheads TCP — position, heading

Designing the conversation

Pressurized gloves make every tap expensive — so voice is ATLAS's primary input. VEGA (Voice Entity for Guiding Astronauts) is built on Rasa, a dedicated conversational-AI platform rather than a keyword recognizer. That distinction matters: astronauts shouldn't have to memorize exact phrases. Trained on sample conversations ("stories"), VEGA interprets diverse natural language — and supports multiple languages, because human spaceflight is a global partnership.

VEGA — One Intent, Many Phrasings Sample dialogue · Rasa NLU
How much water do I have left?
ASTRONAUT → NLU: query_biometrics (H₂O)
Water remaining: 64 percent — approximately 3 hours at current consumption.
VEGA · TTS + VISOR MIRROR
— different phrasing, same intent —
What's the status on my H2O?
ASTRONAUT → NLU: query_biometrics (H₂O)
Water remaining: 64 percent. Would you like consumption trends?
VEGA
Take me to waypoint three.
ASTRONAUT → NLU: navigate (WP-3)
Routing to Waypoint 3 — 240 meters northeast. Path displayed on your minimap.
VEGA → NAVIGATION MODULE

Testing under constraint

As one of two HITL coordinators, I developed the February 2020 usability test plan around the NASA Task Load Index — the standard instrument, invented at NASA Ames in the 1980s, for assessing subjective workload with a human-machine interface. Its six dimensions formed the dependent variables each trial series sought to minimize, and gave us a shared language with the EVA experts reviewing our work.

Then COVID-19 hit. HoloLens deployment was off the table. We adapted: the team tested static designs by displaying them on the HoloLens with a clear background — previewing color and text-size decisions in true AR optics without building in Unity — and converted the moderated test plan into a remote Adobe XD walkthrough paired with a Qualtrics questionnaire.

When COVID canceled in-person testing, adapting the plan became most of the job.

NASA-TLX — The Workload Instrument Illustrative profile of the six dimensions measured per trial
Mental demand
Physical demand
Temporal demand
Performance
Effort
Frustration

The outcome

1st2020 NASA SUITS CHALLENGE
$10KEPIC MEGAGRANT AWARDED
8MODULAR CAPABILITIES SHIPPED IN ATLAS

ATLAS became the foundational system the CLAWS lab continues to build on — VEGA grew into deeper ATLAS and CMCC integration, and the roadmap we authored (RIGEL rock identification, HoloLens 2 eye-tracking, biometric trend prediction suggested by former astronaut Dr. Jim Bagian) seeded the lab's next three years of work.

What I'd do differently: bring geologists in at week one, not month three. Our sampling flow improved more from two hours with a field geologist than from a month of internal iteration.

NEXT CASE STUDY — 03 / 07 Exploration Medical Capability →

CONTENT BASED ON THE CLAWS 2020 NASA SUITS FINAL REPORT. INTERFACE GRAPHICS ARE WEB RECREATIONS, NOT FLIGHT SOFTWARE; TLX PROFILE AND DIALOGUE VALUES ARE ILLUSTRATIVE.