Overview
Biome AI is an interactive spatial computing project designed for the Georgia Tech Library. The project transforms the Crosland Library stairs—a transient, high-traffic area—into a "Digital Biome" that promotes mental wellness. By utilizing computer vision and generative projection mapping, the installation creates a responsive environment that reacts to students' presence with calming, nature-inspired visuals, turning a mundane architectural feature into a site for autogenic relaxation.
Problem
The Georgia Tech Library hosts a massive student population often under high academic stress. While the library offers study spaces, it lacked "passive" restoration zones where students could decompress without actively engaging in a task.
The Prompt: Explore interactive media environments for the library stairs to fascinate, relax, or inspire passersby.
The Core Problem:
How might we utilize the transitional space of a staircase to interrupt the cycle of stress and provide a moment of unsolicited relaxation?
Opportunity Statement
The design process began with a divergent exploration phase, generating seven unique "Opportunity Statements" ranging from gamified group play to career inspiration stories.
We utilized a matrix of impact vs. feasibility to narrow our focus. We selected the Relaxation typology, defined by the following opportunity statement:
"As part of our efforts to make students feel relaxed, the Student Wellness Center will provide a moment of relaxation at the Crosland Library Stairs by having a flower bloom in real-time behind them as they sit & study."
Site - The Stairs
Site Analysis:
Site: Crosland Library Stairs. Affordances: Quiet, communal, physically tiered (ideal for distinct “zones”).
Constraints: Content must be subtle and non-distracting to nearby studying students.
The flow of processes we followed
Mapping out how the journey could look like at the site
The Experience Pillars:
Autogenic Relaxation: Using visual imagery (nature scenes) to trigger a relaxation response.
Guided Imagery: distinct habitats (e.g., underwater, forest) that transport the user away from the academic setting.
Responsive Ecology: Creatures have encoded behaviors (reaction to other animals, reaction to the biome, reaction to the human creator).
Concept Physical Design
Sketching how human sitting would work in a functioning stair space ona given day
3. Conceptual Content Design
Autogenic Relaxation
Guided Imagery
Responsive Ecology
Home Screen
Adding Creature
Creature Builder through Shapes
Prototyping
The development followed a rapid iterative cycle using P5.js for visual generation and physical prototyping for spatial testing.
Sketching a higher fidelity after conceptual content design
Iteration 1: Abstract Behaviors We started with simple geometric circles. We coded collision detection and basic flocking behaviors to ensure the “creatures” felt alive before worrying about their visual fidelity.
Mapping out how the journey could look like at the site
Iteration 2: Creature Definition We replaced shapes with recognizable biological forms to evoke a stronger emotional connection:
Dog: programmed to jump upon gesture triggers.
Bird: flies based on motion detection.
Fish: swims toward the user when they settle into the space.
Tech Stack:
Input: Webcam/Sensors for presence and gesture detection.
Processing: P5.js sketches overlayed on the video feed.
Output: Projectors mapped to the specific geometry of the stair risers.
1. Fish - Swimming to the gesturer on trigger.
2.Dog - Jumping on gesture trigger.
3. Bird - Flying on gesture trigger.
Final Demo at the Site:
Key Findings:
Scale Matters: Creatures needed to be large enough to be seen from a distance but small enough to fit on a single stair riser.
Latency: The “delight” factor relied heavily on the responsiveness of the creature. Any lag broke the immersion.
Lighting: The ambient light of the library required high-contrast visuals for the projection to be legible.