This research introduces a simulation-guided approach for safer and more reliable LLM-generated code in autonomous driving systems.

• Combines LLMs with simulation-based validation to evaluate generated code before deployment
• Focuses on Adaptive Cruise Control (ACC) as a practical use case
• Demonstrates enhanced code reliability through simulation verification compared to standalone LLM code generation
• Provides a framework for reducing development time while maintaining safety standards

This work addresses critical safety-engineering challenges for autonomous vehicles by creating a structured approach to validate AI-generated code before it controls real vehicles on the road.

On Simulation-Guided LLM-based Code Generation for Safe Autonomous Driving Software