This research explores how Foundation Models (FMs) like Large Language Models can enhance autonomous driving systems through improved perception, reasoning, and decision-making capabilities.

• Addresses key challenges in integrating FMs into safety-critical autonomous systems
• Provides a systematic roadmap for FM integration in complex cyber-physical systems
• Examines reliability and safety considerations unique to autonomous driving applications
• Proposes evaluation frameworks for FM-powered autonomous systems

For the aerospace industry, this research offers valuable insights on implementing AI in safety-critical navigation systems that share similar requirements for perception, reasoning, and operation in uncertain environments.

Foundation Models for Autonomous Driving System: An Initial Roadmap