Inspiration
What it does
How we built it
Challenges we ran into
Accomplishments that we're proud of
What we learned
What's next for Seu Amigo Condutor
Inspiration
Modern GPS applications often overwhelm drivers with cluttered 2D/3D maps full of lines, icons, and notifications, causing dangerous visual distractions. Agente Copiloto - Guia Amigo was created to fundamentally change how we navigate: replacing visual strain with intuitive, humanized audio navigation. Instead of counting meters, drivers receive clear guidance based on real-world visual landmarks—just like a helpful friend in the passenger seat.
What it does
Agente Copiloto is a multimodal AI-powered co-driver designed to increase road safety and comfort:
- Humanized Voice Guidance: Directs the driver using landmark references rather than abstract distance units.
- Custom AI Personas: Offers distinct co-driver personalities (e.g., Carlos - Friendly Co-driver, Sgt. Rocha - Tactical Commander, Carol - Upbeat Guide) to suit different driving preferences.
- Computer Vision & Fatigue Monitoring: Connects to the device's camera to analyze driver alertness, detecting signs of fatigue, drowsiness, and lane posture in real time with instant audio alerts.
- Proximity & Risk Zone Alerts: Automatically monitors map data for high-risk areas (theft hotspots, flood zones, deep potholes, dangerous intersections) and lowers audio levels to deliver timely safety cautions.
- Voice-Controlled Navigation: Features a hands-free microphone interface for destination input and route planning.
How we built it
- AI & Logic: Built using Google AI Studio and Gemini API for natural language understanding, persona simulation, and context-aware routing instructions.
- Computer Vision: Implemented real-time video stream analysis via WebRTC/Browser Camera APIs to track driver eye movement and posture for fatigue detection.
- Frontend & UX: Developed a sleek, dark-mode HUD interface optimized for low distraction and night-time driving.
- Deployment: Hosted as a cloud application on Google Cloud Run for fast response times.
Challenges we ran into
Integrating real-time vision algorithms alongside continuous audio synthesis without latency was a major hurdle. Balancing heavy AI inference with a lightweight mobile browser interface required strict optimization to keep performance smooth and responsive while driving.
Accomplishments that we're proud of
- Creating an end-to-end multimodal co-pilot experience combining vision, audio, and geolocation.
- Successfully transforming raw geographic data into natural, humanized directions.
- Designing a safety-first solution that actively keeps the driver's eyes on the road.
What's next for Agente Copiloto
- Vehicle CAN-Bus Integration: Connect directly with car sensors for real-time telemetry (speed, fuel, diagnostics).
- Community Risk Mapping: Enable drivers to report real-time hazards via simple voice commands.
- Offline Audio Cache: Optimize models to run edge inference locally during poor cellular coverage.
Log in or sign up for Devpost to join the conversation.