Skip to content
Vision Pilot v1.2
Autoware Foundation · v1.2 released

Open source L2 ADAS,
from one camera.

A productionizable and safety-certifiable L2 ADAS stack — ACC, AEB, LKAS, LDW, ISA and single-lane highway autopilot — running on a single front-facing camera, with no HD maps. Complete codebase and AI model weights, free under Apache 2.0.

Apache 2.0 licence Autoware Discord GitHub stars Latest release
1camera, no HD map
10 Hzdesign cadence
3–5INT8 TOPs
Apache 2.0weights included
Capabilities

Entry-level L2, done properly

Vision Pilot targets the in-lane autonomous driving feature set that automotive OEMs and Tier-1 suppliers ship in series production passenger cars.

ACCAutonomous cruise controlLongitudinal control that holds a safe gap to the closest in-path object.
FCWForward collision warningAlerts the driver when a collision with the lead object becomes imminent.
AEBAutonomous emergency brakingApplies braking automatically when the warning is not acted on.
LKASLane keep assistLateral control that centres the vehicle on the estimated ego path.
LDWLane departure warningFlags unsignalled drift across a lane boundary, left or right.
ISAIntelligent speed assistMap-based — keeps the commanded speed within the applicable limit.
LDALane departure avoidanceSteers back into lane rather than only warning about the drift.
ALKAutonomous lane keepSustained in-lane lateral control on top of LKAS.
AutopilotSingle-lane hands-free highway autopilotCombined longitudinal and lateral control for highway driving, built on the feature set above.
Vision Pilot capability summary: one front-facing 2 MP camera at 50 degrees horizontal field of view, the supported ADAS feature set, and a 3–5 INT8 TOPs compute budget
Vision Pilot at a glance — sensing, supported features and compute budget.
Hands-free autopilot has a defined operational domain. It is available only on highways and motorways with clearly marked lane lines, in fair weather and visibility, across the full range of highway speeds (0–70 mph), and away from construction zones and roadwork objects. A human driver must monitor and supervise the system at all times.
Architecture

Hybrid end-to-end AI

Camera frames are processed in parallel by perception models for safety and by an end-to-end model for performance. A Safety Guardian fuses both paths before anything reaches the planner.

Vision Pilot architecture: camera input feeding parallel perception and end-to-end AI branches, fused by the Safety Guardian into longitudinal and lateral planning
The Vision Pilot processing pipeline, from camera frame to actuation command.

No HD maps

Vision Pilot operates fully mapless, following the road in real time. No localisation stack, no map maintenance, no geofence to keep current.

One monocular camera

A single front-facing RGB camera with 50–55° horizontal FoV at 2 MP. That is the entire sensor specification.

Read the architecture guide →

Quick start

Three ways to run it

Sample OpenLane data is provided so you can see the full pipeline in action before touching a vehicle or a camera.

Best if you intend to modify the stack. Requires ONNX Runtime.

git clone https://github.com/autowarefoundation/vision_pilot.git
cd vision_pilot/VisionPilot
mkdir build && cd build

cmake -DONNXRUNTIME_ROOT=<ONNX_RUNTIME_ROOT_PATH> ../
make -j$(nproc)

./VisionPilot

Add -DENABLE_ROS2_INTERFACE=ON for ROS 2, or -DENABLE_OCCUPANCY=ON for the bird's-eye Occupancy window.

In action

Open-loop and closed-loop

Replay recorded drives for evaluation, or close the loop in CARLA to test how the stack actually behaves when its own commands move the vehicle.

Object detection, closest-in-path distance, ego path, lane departure and speed limit, rendered live alongside the heuristic Occupancy bird's-eye panel.

The same binary, fed by a ROS 2 bridge from CARLA, steering and braking a virtual vehicle. See the simulation guide.

Releases

Where the project is

Tagged releases land on GitHub with source, Debian packages and model weights.

v1.2Latest10 Aug 2026
  • Occupancy bird's-eye view, with configurable colours
  • Rerun logging: stream per-frame data straight to an .rrd recording
v1.127 Jul 2026
  • Optional Occupancy BEV window beside the HUD
  • Camera calibration guide and homography tooling
v1.0Milestone6 Jul 2026
  • First full L2 feature set on the hybrid end-to-end architecture
  • Debian packaging, Docker images, CARLA closed-loop support
v0.92 Mar 2026
v0.520 Jan 2026
Roadmap

What is coming next

  • Fusion between the front-facing camera and automotive RADAR
  • 8 MP camera support at 120° horizontal field of view
  • Safety verification and standards compliance (ISO 26262, ISO 8800)
  • SODA.Sim simulator integration

Full release history →

Community

Built in the open, weekly

Vision Pilot is developed by the Privately Owned Vehicle working group of the Autoware Foundation. Everything happens in public.

Not a production system out of the box. Vision Pilot is designed to be productionizable and safety-certifiable, but any deployment on a real vehicle is your responsibility. Read the disclaimer and the functional safety documentation first.