SafeAuto presents a novel approach that unifies perception and reasoning for autonomous vehicles using Multimodal Large Language Models (MLLMs) with embedded safety knowledge.

• Bridges the gap between high-level reasoning and low-level vehicle control
• Enhances driving safety through knowledge-rich multimodal AI systems
• Addresses key limitations in traditional autonomous driving systems
• Demonstrates how foundation models can be specialized for critical safety applications

This research represents a significant engineering advancement by creating a framework where AI can simultaneously process visual data and apply safety reasoning - potentially reducing accidents and improving autonomous vehicle reliability in complex driving scenarios.

SafeAuto: Knowledge-Enhanced Safe Autonomous Driving with Multimodal Foundation Models