BeamNG/BeamNGpy

What it solves

BeamNGpy provides a Python API for BeamNG.tech, allowing researchers and developers to programmatically control a high-fidelity vehicle simulation. It eliminates the need for manual interaction with the game engine, enabling automated testing, AI training, and complex scenario generation.

How it works

The library acts as a client that communicates with the BeamNG.tech server. It allows users to define scenarios, spawn vehicles, and send low-level control inputs (like steering and throttle) or high-level AI commands (such as following a trajectory or spanning a map). It also provides access to simulated sensor data from cameras, Lidars, and IMUs, which can be configured to include noise for realism.

Who it’s for

It is designed for academia and industry professionals working on autonomous driving, vehicle dynamics, and robotics research.

Highlights

• Remote Vehicle Control: Individual control over steering, throttle, gear shifting, and lights.
• AI Integration: Ability to configure the simulator's built-in AI to follow waypoints or other vehicles.
• Dynamic Sensor Models: Support for color/depth cameras, semantic annotations, Lidars, and ultrasonic sensors.
• Simulation State Access: Access to the road network geometry and active scene objects for analysis.
• Multi-Client Support: Multiple Python processes can connect to a single simulation to control different vehicles independently.