GM Develops Low-Velocity Maneuvering Capability in Self-Driving Vehicles Using MATLAB and Simulink
“MATLAB and Simulink enabled us to construct and test the autonomous vehicle inside a 3D world.”
Key Outcomes
- Accurately simulated a test vehicle’s autonomous maneuverability over the last 100 meters when GPS is unavailable
- Connected vehicle systems to navigate the environment using one platform between proof of concept and production environments
- Accelerated development and improved architecture management
The last 100 meters of an autonomous vehicle’s trip can entail unique challenges. There is variability in direction and often no road markings or GPS. For General Motors, using MATLAB® and Simulink® in the development of autonomous-driving vehicles is key to solving these challenges.
Using Simulink, GM engineers constructed a 3D, virtual world to simulate the actual test vehicle’s path and performance, including low-velocity maneuvering. For photorealism in the simulation, they generated computer vision algorithms in MATLAB.
One of the major steps forward for GM was the ability to replicate the vehicle’s fisheye images in the 3D model using MathWorks products with third-party tools.
GM uses Simulink for the Model-Based Design, CarSim for vehicle dynamics, Unreal Engine for image rendering, ROS2 for visualization and debugging, and dSPACE for real-time validation.
With the progress GM has made using these development tools, the company hopes to be selling autonomous-driving vehicles by mid-decade.
