Angry Sheep

Tabletop museum exhibit: children collaborate to guide digital sheep into a pen using physical tools. Built for the Children's Museum of Pittsburgh, targeting ages 6–8.

A circular table displays a projected scene with digital sheep. Children place physical blocks, sheepdogs, and grass on the table to herd the sheep — but rough handling triggers crisis mode, and only calming actions (feeding, petting, kind words) can help.

Repo layout

Path	Purpose
web/	Frontend web app (p5.js + Vite) — sheep simulation, game logic, session flow
server/	Python ArUco marker tracking server (WebSocket)
markers/	Printable ArUco marker assets (see markers/README.md)
(root)	Docs, design concept, roadmap, session logs

Web app (frontend)

cd web
npm install
npm run dev

Opens at http://localhost:5173/. The app runs entirely with mock input (mouse/keyboard) — no physical hardware needed.

Controls (mock input)

Action	How
Move a tool	Drag it
Rotate a tool	Scroll wheel while hovering
Pet a sheep	Hold left-click near a sheep
Speak kindly	Hold V key

What's implemented

• Scene — Circular play area with black mask (projection-ready), circular pen with openings
• Sheep — Wandering, flocking, tool reactions (flee dogs, attract to grass, bounce off blocks)
• Crisis mode — Stress builds from sheepdogs, sheep turn red/angry, ignore pushing, can multiply. Mad sheep stay mad until a player intervenes — no passive decay.
• De-escalation — Grass calming (mad sheep seek grass for comfort), petting (heart animation), voice input
• Grazing — Calm sheep are briefly attracted to grass but fill up quickly and wander away. Mad sheep stay hungry and are always drawn to grass.
• Session flow — Intro animation → 3-min timer → win celebration or timeout → auto-reset
• Hints — Icon-based hint bubbles appear on unresolved crisis sheep
• Tuning panel — Click "Tune" (top-right) to adjust simulation parameters at runtime, organized by category (Sheep, Grass, Tools). Save/load named presets (persisted in localStorage).

Architecture

The frontend consumes a shared JSON contract:

{
  "tools": [{ "type": "block", "id": 0, "x": 0.45, "y": 0.30, "angle_deg": 90 }],
  "voice": { "active": false, "sentiment": null },
  "pet":   { "active": false, "x": null, "y": null }
}

Currently produced by web/src/input.js (mock). In production, a WebSocket connection to server.py replaces it — the rest of the app doesn't change.

Python server (tracking layer)

cd server
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
python server.py

Uses the default camera and serves WebSocket on ws://127.0.0.1:8765. Streams ArUco marker positions as JSON.

Team

• Diane Hu, Cameron, Nevil
• CMU MDes/MPS Studio

Docs

• Design concept — Research foundation, design principles, emotional mechanics
• Roadmap — Two-track development plan (frontend + physical layer)
• Usability test plan — Round 1 testing protocol for the digital prototype
• Session log — 2026-03-28 — Build progress: A1–A4 (scene, sim, crisis, session flow)
• Session log — 2026-03-28 #2 — Tuning panel, grazing behavior, grass mechanics rework