HomeMy WorkLinkedInResume

The CDASim Project & CARMA-CARLA ROS 2 Integration

A full-stack ROS 2 integration for the CARLA autonomous vehicle simulator and connected systems.

GitHub Page
Carla UE5 Wiki

About

The CARMA-CARLA Integration Bridge is a critical software component that enables high-fidelity co-simulation for the U.S. Department of Transportation's CARMA Platform. It connects the world-class vehicle dynamics of the CARLA simulator with the cooperative driving automation logic of CARMA, allowing for robust, scalable testing of autonomous vehicle behaviors.

Working for Leidos through the University of Georgia's MSC Lab, my primary role was to spearhead the migration of this entire integration stack from ROS 1 to ROS 2, refactoring the codebase, overhauling the build system, and engineering custom solutions to restore critical functionalities that were deprecated in newer versions of the CARLA simulator. As a CS Junior at UGA, this was a great learning experience.

A key deliverable of my work was the development of a custom ROS 2 bridge from the ground up. This implementation successfully re-established the full vehicle and sensor transform framework, and is now pending review for potential adoption by the official CARLA team in Spain. I also developed my own fork of the bridge which you can find here: https://github.com/willv678/carla-ros2-bridge.

Technical Deep Dive

This project involved a full-stack effort, from low-level vehicle control logic to high-level system architecture and deployment. I successfully resolved over 20 complex engineering tickets, representing more than 75 story points of development work.

Key Contributions

  • ROS 1 to ROS 2 Migration: I converted the entire ROS 1 codebase, which included 15+ nodes, to ROS 2 (Humble). This involved updating the build system from catkin to colcon, converting XML launch files to modern Python launch scripts, and refactoring all node communication to use the rclpy client library.

  • Custom ROS 2 Bridge Development: When the native transform publisher was removed from the official CARLA simulator, I architected and developed a new ROS 2 bridge from scratch. This component is now responsible for publishing the critical map -> vehicle -> sensor transforms, which are essential for the CARMA Platform's localization and route-planning systems to function.

  • Vehicle Control & PID Tuning: A major challenge was debugging the vehicle control pipeline. After connecting all the components, the vehicle would not respond correctly to commands. I diagnosed and fixed the issue by tuning the PID controller parameters in the ackermann_control node and resolving coordinate system mismatches between ROS and CARLA, enabling the vehicle to accurately follow complex trajectories.

CARLA Simulation In Action

A successful test run in CARLA, with the vehicle correctly following it's route given my the CARMA manager running in the background.

  • Full-Stack Debugging: I was responsible for identifying and fixing bugs across the entire simulation stack. This involved a methodical process of isolating components, verifying topic and service connections with command-line tools (ros2 topic echo/info), and injecting debug logs to trace data flow from the high-level CARMA guidance nodes all the way down to the final control commands being sent to the simulator.

Closing Thoughts

This project was an incredible deep dive into the real-world challenges of integrating complex robotics systems. Moving a tool of this scale from ROS 1 to ROS 2 was more than just a code update; it was a complete architectural refactoring that required a deep understanding of every component in the stack. I'm proud to have contributed code that is now a fundamental part of a national research platform, helping to test and validate the autonomous vehicle technologies that will define the future of transportation. Thanks for reading!