LumenRoute

Built for UGAHacks 11 | Theme: Magic

About The Project

LumenRoute transforms existing city infrastructure into an intelligent safety network. We built this platform during UGAHacks 11 to address two critical issues for Georgia commuters: traffic safety and environmental impact.

By leveraging Google Gemini 1.5, we analyze live feeds from over 50 GDOT traffic cameras in real-time. This allows us to detect hazards like flooding, accidents, and debris faster than traditional reporting methods. We combine this visual data with environmental metrics to offer drivers "Eco-Routes"—paths optimized not just for speed, but for lower carbon emissions and higher safety scores.

Key Features

• Guardian Dashboard: Real-time monitoring of Atlanta's traffic network. The system automatically flags hazards using computer vision on live camera feeds.
• Eco-Route Planner: Intelligent routing algorithm that factors in CO2 emissions, air quality, and active hazard data to suggest safer alternatives.
• Scout Mode: A community reporting tool where users can submit infrastructure issues (like potholes or broken chargers). Submissions are verified instantly by Gemini Vision to filter spam.
• Multimodal Analysis: The core engine integrates visual data from cameras, text reports from users, and sensor data (weather/AQI) to make complex routing decisions.

Team Members

• Anish Banda
• Deep Sahu
• Kedarnath Mohan
• Shiv Pandey

Tech Stack

We prioritized performance and real-time capabilities for this build.

Frontend

• React + Vite: For a responsive, client-side application.
• TypeScript: Ensuring type safety across complex data structures.
• Leaflet: Lightweight mapping library for rendering routes and camera clusters.
• Tailwind CSS: Rapid UI development with a consistent design system.

Backend

• Bun: Chosen for its high-performance runtime and built-in tooling.
• Hono: A lightweight web framework optimized for edge deployment.
• Google Generative AI SDK: Powering both the vision analysis (Gemini 1.5 Flash) and reasoning engine (Gemini Pro).

Challenges & Solutions

Real-Time Data Synchronization

Integrating GDOT's ArcGIS MapServer presented a challenge due to the sheer volume of data and rate limits. Querying the full dataset frequently would cause latency. Solution: We implemented a diffing strategy in our backend. The server polls the API at set intervals but only processes and broadcasts updates for cameras whose status has changed, significantly reducing overhead.

Validating User Reports

Crowdsourced data is valuable but prone to noise. We needed a way to verify reports without manual moderation during the hackathon. Solution: We used Gemini Vision as an automated moderator. When a user uploads a photo of a "pothole," the AI analyzes the image context. If it doesn't match the report category, the submission is automatically rejected.

Getting Started

Prerequisites

• Node.js (v18+) or Bun

Installation

1. Clone the repository:

   git clone https://github.com/your-username/lumenroute-ai.git
   cd lumenroute-ai

2. Configure environment variables: Copy config/env.example to .env in the project root and add your API keys (Google Maps, Gemini, etc.).

   cp config/env.example .env

3. Install dependencies and run:

   bun install
   bun run dev

4. Access the application:

   • Frontend: http://localhost:8000
   • Backend: http://localhost:3000

License

Distributed under the MIT License.

Acknowledgments

• UGAHacks for hosting.
• GDOT for providing open access to traffic camera data.
• Google for the Gemini API.
• State Farm and Cox Automotive for their support and resources.