LGSVL

LGSVL, short for LG Silicon Valley Laboratory Simulator, is an open-source autonomous vehicle simulation platform developed by LG Electronics’ Silicon Valley Laboratory. It provides a realistic, controllable environment for developing, testing, and validating self-driving software, enabling researchers and engineers to simulate perception, localization, planning, and control workflows without real-world risk.

The simulator is built on the Unity engine and runs on Windows and Linux. It offers high-fidelity urban scenes with weather, lighting, and traffic dynamics. It supports authentic sensor models for ego and other vehicles, including cameras, LiDAR, and radar, as well as physics-based vehicle dynamics and collision detection. Maps can be authored in OpenDrive or imported as HD maps, and scenes can be augmented with scripting and scenario editing. It includes a library of NPC traffic participants and pedestrians to create realistic traffic flows.

LGSVL provides integration with major autonomous driving stacks and developer tools. It includes bridges to ROS and ROS 2, enabling data exchange with perception and planning modules. It has sample pipelines for Autoware, Apollo, and other platforms, and can export ground-truth data and sensor streams for training and benchmarking. The platform supports data logging, replay, and debugging workflows, and can be extended with custom sensors and behavior models.

First released by LG Electronics in the late 2010s as SVL Simulator and later renamed LGSVL Simulator, the project is maintained as an open-source resource used by universities, researchers, and industry groups for autonomous driving research and product validation. It is widely documented on GitHub and the project site, with ongoing community contributions and updates.