This research introduces a novel framework that overcomes the speed limitations of LLMs in autonomous vehicles by using parallel processing to enhance real-time decision-making and interaction.

• Combines a fast-acting Actor module for immediate decisions with a deliberative Reasoner powered by LLMs
• Implements memory retrieval to provide context and improve driving decisions
• Demonstrates significant improvements in vehicle-to-vehicle communication and intention signaling
• Addresses critical safety concerns through better prediction of human driver behavior

This engineering breakthrough creates a path toward safer autonomous vehicles that can better communicate intentions to human drivers, potentially accelerating commercial AV adoption.

Interact, Instruct to Improve: A LLM-Driven Parallel Actor-Reasoner Framework for Enhancing Autonomous Vehicle Interactions