Researchers present Large Language Model framework for interactive decision-making in autonomous driving within mixed traffic scenarios

A framework using Large Language Models (AI systems that understand and generate text) augments autonomous vehicle decision-making by modeling multi-vehicle scenes semantically, parsing intent of surrounding agents, and selecting candidate maneuvers under jointly enforced safety and efficiency constraints.

The approach translates final driving decisions into natural-language messages broadcast through an external Human-Machine Interface (a communication system between vehicle and nearby road users), completing a loop from scene understanding to action to language.

Experiments in a cluster driving simulator demonstrate the method outperforms traditional baselines across safety, comfort, and efficiency metrics, with evaluation indicating a high degree of human-likeness in decision making.