Mei-Chu Hackathon 2025 · Maker Track · 🥈 2nd Place
An AI-powered traffic signal coordination system for Hsinchu County, featuring GLIDE green-wave optimization, Transit Signal Priority (TSP), and anti-bunching control.
GLIDE-Lite is a browser-based traffic simulation and visualization system that demonstrates how intelligent signal coordination can improve corridor traffic flow. Built during a 24-hour hackathon using Hsinchu County open data, the system simulates a 5-intersection arterial corridor and compares three control strategies in real time.
The project was recognized by the Hsinchu County Transportation Bureau and Youth Department Director for its potential public value.
Three simulation modes:
- Fixed Timing — baseline with pre-timed signals
- GLIDE Green Wave — progressive signal coordination that creates a "green wave" for vehicles traveling at the target cruise speed, reducing stops
- GLIDE + TSP — adds Transit Signal Priority for buses, with green extension, phase hold, and anti-bunching logic to maintain service regularity
Real-time visualization:
- Animated multi-lane corridor with 5 intersections modeled after real Hsinchu arterial roads (大學路口、光復路口、中正路口、民族路口、竹北側路口)
- Per-vehicle rendering with lane assignment, bus stop dwell, and stopped-vehicle indicators
- Live KPI dashboard showing progression rate, average delay, stop count, and travel time
Bus operations:
- Configurable headway, dwell time, and tolerance range
- TSP parameters: max green extension, max hold duration, cooldown period
- Bunching simulation toggle and anti-bunching control
- Per-route and per-stop monitoring tables
| Layer | Tools |
|---|---|
| Frontend | React 18, Vite 5, TailwindCSS 3 |
| Visualization | HTML5 Canvas (custom renderer) |
| Simulation engine | Python backend (FastAPI at localhost:8001) |
| RL training | SUMO-RL (reinforcement learning for multi-intersection coordination) |
| Dev tooling | ESLint, Prettier, Vitest, Testing Library |
# Clone the repo
git clone https://github.com/ellatso/2025meichuhackathonteam8.git
cd 2025meichuhackathonteam8
# Install dependencies
npm install
# Start the frontend dev server
npm run devNote: The simulation backend (FastAPI server on port 8001) must be running separately for the simulation to work. The frontend sends POST requests to
/glide/simwith configuration parameters.
2025meichuhackathonteam8/
├── public/ # Static assets
├── src/ # React source code
├── index.html # Main simulation UI (Canvas + controls)
├── package.json # Dependencies and scripts
├── vite.config.js # Vite configuration
├── tailwind.config.js # TailwindCSS configuration
└── postcss.config.js # PostCSS configuration
| Parameter | Default | Range | Description |
|---|---|---|---|
| Cycle length | 90s | 60–180s | Signal timing cycle |
| Cruise speed | 50 km/h | 30–70 km/h | Target green wave speed |
| Simulation steps | 300 | 180–600 | Duration of simulation run |
| Traffic volume | 1700 vph | 600–2200 vph | Per-direction, per-lane flow rate |
| Bus headway | 6 min | 2–20 min | Scheduled departure interval |
| TSP max extension | 10s | 0–30s | Max green time added for bus |
| TSP max hold | 30s | 0–60s | Max red hold for bus approach |
| TSP cooldown | 120s | 0–600s | Min gap between TSP activations |
The system reports the following performance indicators after each simulation run:
- Progression rate — percentage of vehicles that pass through the corridor without stopping
- Average stops per vehicle — mean number of red-light stops on the mainline
- Average signal delay — mean time lost waiting at signals
- Average travel time — end-to-end corridor traversal time
- TSP / Hold events — count of transit priority activations
In our hackathon demonstration, the GLIDE green wave mode achieved approximately 30% improvement in simulated corridor flow compared to the fixed-timing baseline, while the TSP mode maintained bus schedule adherence without significantly degrading general traffic performance.
- Hsinchu County Government — open traffic data
- Mei-Chu Hackathon 2025 organizers
- SUMO (Simulation of Urban MObility) — traffic simulation framework
MIT