ucla-mobility/OpenCDA

What it solves

OpenCDA addresses the need for an open-source research and engineering framework to develop and evaluate Cooperative Driving Automation (CDA). It provides a standardized environment to test how multiple connected and automated vehicles (CAVs) can cooperate—such as through platooning or shared perception—rather than focusing on a single vehicle in isolation.

How it works

The framework operates as a full-stack simulation platform written in Python. It integrates two primary simulation tools: CARLA (for high-fidelity vehicle dynamics and environment) and SUMO (for traffic simulation). OpenCDA layers a cooperative driving system on top of these, including modules for perception, localization, planning, control, and V2X (Vehicle-to-Everything) communication. It also includes a data manager, a scenario manager, and a library of research pipelines for implementing specific CDA algorithms.

Who it’s for

It is designed for researchers and engineers in the field of intelligent transportation systems and automated driving who are conducting early-stage fundamental research on cooperative vehicle behaviors.

Highlights

• Full-Stack Python Implementation: Includes integrated modules for the entire driving pipeline from perception to control.
• Co-Simulation: Seamlessly combines CARLA and SUMO to balance environmental detail with traffic scale.
• V2X Support: Specifically built to simulate communication and cooperation between multiple vehicles.
• Benchmark Suite: Provides baseline maps, state-of-the-art algorithms, and evaluation metrics for standardized testing.
• Research Pipelines: Includes open-source code for advanced modules like cooperative perception and platooning.