PROJECT ARES

AI-Adaptive Lunar Habitat Design

Psychological Wellbeing Through Intelligent Environments

8-Week Team Project

Teams of 3–4

200 Points Total

Psychology · Engineering · AI · Ethics

A student design challenge in space psychology and AI systems

The problem

Structural engineering on the Moon is largely solved. Crew psychology is not.

18

months

Mission duration

— same as longest

ISS stays

6

crew

People sharing

a space smaller

than a house

#1

risk

Behavioural health

ranked top mission

risk by NASA

0

outside

No parks, no fresh

air, no spontaneous

human contact

Your mission

Design the psychological support architecture for a six-person lunar habitat on an 18-month mission. Your system must use AI, biosensors, voice analysis, and facial recognition to detect and respond to crew psychological states — and adapt the living environment in real time.

Detect

Biosensors, voice prosody, facial recognition, and behavioural tracking feed continuous data to the AI

Analyse

AI cross-references signal streams, triages severity, and builds a longitudinal baseline for each crew member

Respond

The habitat adapts — lighting, audio, temperature, scent, schedules, and social nudges — automatically and proportionally

Six conditions to address

Your system must detect and respond to all six.

Anxiety

Acute stress from confinement, pressure, or safety fears

Depression

Sustained low mood and withdrawal building over days

Interpersonal conflict

Crew tension, avoidance, and communication breakdown

Cognitive fatigue

Declining task performance from sustained high-stakes work

Autonomy deprivation

Loss of personal agency in a tightly scheduled environment

Sensory monotony

Perceptual habituation and boredom from unchanging surroundings

Sensor architecture

Phase 1 (Weeks 1–2): Research which sensors detect which conditions

HRV

Heart rate variability

Anxiety · stress · fatigue

GSR

Galvanic skin response

Acute anxiety · arousal

Sleep

Sleep architecture tracker

Depression · fatigue · circadian

Voice

Voice prosody analysis

Mood · conflict · fatigue

FACS

Facial action coding

Emotion · depression · conflict

Motion

Movement / proximity

Isolation · conflict · autonomy

Cortisol

Cortisol wearable

Chronic stress · depression

Air

CO₂ / air quality

Cognitive fatigue · irritability

Key design question

Every sensor added to a lunar habitat costs mass and power. Justify why each sensor is worth its weight. What happens when one fails? How does your system gracefully degrade?

Habitat spatial design

Phase 2 (Weeks 3–4): Design zones for psychological purpose, not just function

Communal zone

Shared meals, conversation, group activities. Warm lighting to draw crew together.

Private retreat

Acoustically shielded. Each crew member's personal space for rest and decompression.

Work & mission bay

Cool 6500K lighting, structured airflow. Optimised for sustained cognitive performance.

Exercise bay

Energising light and audio auto-activate. Screens show Earth views. Movement is medicine.

Nature / growth bay

Plant growth area. Green light spectra, petrichor scent. Behavioural activation for depression.

Transition corridor

Lighting and temperature shift between zones create psychological cues for mode-switching.

Deliverable: Annotated floor plans for two configurations — nominal operations and conflict protocol

Room Design: Communal vs. Private Space

ENGINEERING DESIGN PROCESS

Define → Specify → Design → Test → Evaluate

Room 1: Communal Hub

PURPOSE

Shared meals, group conversation, leisure, and mission debriefs. The social anchor of the habitat — where crew cohesion is built and maintained over 18 months.

MANAGING THIS SPACE

• Scheduling: AI coordinates shared meal windows; prevents forced overlap for crew in conflict
• Ambient cues: Warm 2800K lighting and low-level scent draw crew without compulsion
• Conflict protocol: Acoustic zones create separation; spatial rearrangement reduces forced proximity
• Shared ownership: No single crew member controls this space; all design decisions must be negotiated

Design task: Scaled floor plan with dimensions, sensor placements, furniture layout, and lighting/audio zones. Justify every element.

KEY QUESTIONS

• How do you make participation feel chosen, not compelled?
• What minimum floor area prevents claustrophobia for six people sharing one room?
• How does the layout change during a conflict protocol?
• Where are sensors placed, and which crew members know they are active?
• How does the AI draw people together without making anyone feel watched or managed?

Room 2: Private Retreat Pod

PURPOSE

Individual decompression, sleep, privacy, and personal agency. The one space each crew member controls completely — including whether the AI has any access at all.

THE KEY DISTINCTION

• AI access: No cameras inside. Environmental controls only on crew member request. The AI monitors the door, not the room.
• Individual control: Lighting, temperature, and audio set entirely by the occupant. No AI override.
• Physical boundary: Lockable. No crew member may enter without permission. Design the lock mechanism.
• Emergency override: Define the exact conditions — if any — under which the AI or CMO may override this privacy boundary.

Design task: Scaled floor plan with ergonomic sleep space, personal storage, sensory controls, and the privacy boundary. Specify dimensions.

KEY QUESTIONS

• What minimum volume allows a human to sleep and decompress without feeling sealed in?
• How does the AI detect distress if it cannot see or hear inside the room?
• What personal items does each crew member pre-load into their pod before launch?
• How does acoustic isolation work without making the occupant feel cut off from the crew?
• Under what exact conditions may the CMO or AI enter or monitor this space?

Deliverable: Scaled floor plan for each room • Sensor placement map • Written justification for every design decision • Specification of the communal–private boundary and how the AI manages the transition

AI response system design

Phase 3 (Weeks 5–6): Design complete response protocols for each condition

Mild / early

Auto — silent background adjustment. No crew notification.

Moderate

Active intervention + crew notification. Options offered.

Severe

Immediate escalation to Chief Medical Officer.

Environmental levers available to the AI

Lighting

CCT · lux · circadian cycle · blue-light filter

Audio

Binaural beats · soundscapes · acoustic zones

Thermal

Temperature · airflow · radiant warmth

Scent

Lavender · citrus · petrichor · ion levels

Routine

Schedule · activity prompts · breaks

Social

Earth calls · shared meals · AI companion

Visual

Home videos · nature scenes · Earth views

Example: Anxiety response flowchart

Students must produce flowcharts for at least 3 of the 6 conditions — here's the structure to follow

Signal detected

HRV drop · elevated HR · tense vocal tone · disrupted sleep

Severity triage

Score 4 signal streams · weight by duration · mild / moderate / severe

Auto intervention

Dim to 2400K amber · alpha binaural beats · raise temp to 23°C · lavender scent

Monitor 20 min

Track HRV · HR · vocal tone · movement patterns

Improving?

Yes → restore gradually · log event

No → escalate to CMO · crew check-in

Speed matters for anxiety

Anxiety spikes fast. The AI must act within minutes. Monitor window is 20 min — not 20 hours.

Silent first, then escalate

Most calming interventions work better when the crew member isn't aware they're happening.

Never override consent

If a crew member explicitly refuses an intervention at any tier, the AI must stop and log.

Ethics & hard limits

Phase 4 (Weeks 7–8): The hardest design questions have no easy answers

Surveillance vs. privacy

The AI must watch continuously to help — but crew members live in the habitat 24/7. Where are the boundaries?

Autonomy vs. intervention

If a crew member refuses help, should the AI back off even if the crew's safety might be at risk?

Transparency vs. distraction

Should the AI explain what it's doing in real time — or would constant notifications themselves cause anxiety?

Individual vs. team

Crew member A's data reveals a risk. Can the AI act in a way that protects others without revealing A's private state?

Your ethics report must define hard limits: what the AI is NEVER permitted to do without human authorisation.

Deliverables & grading

Required deliverables

Sensor-to-condition detection matrix

Annotated floor plans — 2 configurations

AI response flowcharts for 3+ conditions

Ethics & limits report (800–1200 words)

Agency review board presentation (15 min)

Individual reflection journal — 3 entries

Tabletop simulation — 20 pts

Interface mockup — 20 pts

★ HIGH-PRIORITY FOR DESIGN ENGINEERING: Tabletop simulation and interface mockup are full rubric deliverables. In engineering design, testing under failure conditions and designing for the user are not extras — they are the work.

★ Tabletop simulation

20 pts

★ Interface mockup

20 pts

Points breakdown (240 total)

AI response protocols

50 pts

Ethics & limits

40 pts

Sensor architecture

35 pts

Spatial design

35 pts

Presentation

25 pts

Reflection journals

15 pts

8-week timeline

Wks 1–2

Research & sensors

ICE / Antarctic mission literature review

Sensor-to-condition matrix

Identify 2 failure modes per sensor

Wks 3–4

Spatial design

Floor plan with 6 defined zones

Sensor placement rationale

Two-configuration comparison

Wks 5–6

AI response design

Response protocols for all 6 conditions

Flowcharts for 3+ conditions

Consent & crew notification architecture

Wks 7–8

Ethics & present

Ethics report — 4+ tensions addressed

Hard limits defined

Review board presentation

Checkpoints: Week 2 reflection journal due · Week 5 reflection journal due · Week 8 presentation + final submission + journal 3

Final review board

Week 8: Present to the agency

Defend every sensor choice on mass, power, and detection accuracy

Explain your AI severity triage logic under adversarial questioning

Walk the board through a live scenario: two conditions, simultaneously

Justify your hard limits — what the AI will never do, and why

15 minutes presentation · 10 minutes Q&A from the review panel

Good luck — the Moon is waiting.