Autonomous Car Deployment
Full-stack behavior planning and real-world maneuver execution on a Kia Soul platform — from simulation to road testing.
Designed and deployed a complete behavior planning pipeline on a Kia Soul autonomous vehicle platform. The system integrates perception, prediction, and planning to execute real-world driving maneuvers in dynamic traffic scenarios.
Key contributions:
- Developed a prediction-aware behavior planner combining task-and-motion planning (TAMP) with QCNet-based multi-agent trajectory prediction
- Implemented real-time decision-making under safety constraints for lane changes, intersection handling, and obstacle avoidance
- Validated the full pipeline end-to-end: from CARLA simulation through hardware-in-the-loop testing to on-road deployment
Stack: ROS, Apollo Auto, CARLA, C++, Python