autowarefoundation/vision_pilot
Free and fully open-source L2 ADAS stack powered by End-to-End AI technology
What it solves
Vision Pilot provides a production-ready, open-source software stack for Level 2 (L2) Advanced Driver Assistance Systems (ADAS). It enables passenger cars, buses, and trucks to perform in-lane autonomous driving tasks without relying on expensive 3D high-definition maps.
How it works
The system uses a "Hybrid End-to-End AI Architecture" that processes data in parallel using two different AI approaches for a balance of safety and performance. It relies on a single front-facing monocular camera and integrates three specialized AI models:
- AutoSpeed: Detects the closest object in the vehicle's path.
- AutoSteer: Detects future waypoints for the vehicle's path.
- AutoDrive: Estimates road curvature and detects the presence and distance of in-path objects.
Who it’s for
It is designed for automotive OEMs and Tier-1 suppliers for integration into series production vehicles, as well as researchers interested in open-source ADAS technology.
Highlights
- Mapless Operation: Operates in real-time without the need for 3D HD maps.
- L2 Feature Set: Supports ACC (cruise control), AEB (emergency braking), LKAS (lane keep assist), and a single-lane hands-free highway autopilot.
- Hardware Efficient: Designed to run with a standard 1MP-2MP monocular camera.
- Open Weights: Provides both the codebase and AI model weights under the Apache 2.0 license.