LLM4AD: Large language models for autonomous driving

By Ziran Wang on 01/30/2025 from 2:00 p.m. to 3:00 p.m. in 1611 Titan Dr., Rantoul, IL 61866

Join Ziran Wang of Purdue University as he presents in person at the Spring 2025 Kent Seminar Series Thursday, January 30, from 2-3 p.m. (CT).

The Spring 2025 semester is set to feature 14 presentations, each addressing a topic related to autonomy in transportation. See the full lineup of speakers for Spring 2025 semester.

Pizza and soft drinks will be provided beginning at 1:30 p.m. in the ICT Classroom. 

All presentations will be held on Zoom, but some speakers will present in person at ICT.

Join Zoom Meeting
https://illinois.zoom.us/j/89890781073?pwd=CewiD3535GNiWvliWpS6nqBksMqnAE.1 

Meeting ID: 898 9078 1073
Passcode: 116680

Abstract and Bio

Integrating large language models into autonomous driving technology has transformative potential in perception, decision-making and human-vehicle interaction. This talk introduces LLM4AD, a framework leveraging LLMs to improve autonomous driving systems through natural language understanding, situational reasoning and personalized motion control. The system features a novel benchmark for evaluating instruction-following capabilities and employs a retrieval-augmented generation-based memory module for continuous learning, enabling adaptive control strategies informed by human feedback. Field experiments using a drive-by-wire-enabled autonomous vehicle demonstrate the system’s ability to translate complex natural language instructions and environmental inputs into actionable control policies.

Wang is an assistant professor at Purdue University’s Lyles School of Civil and Construction Engineering and assistant director of the Institute for Control, Optimization and Networks. His research focuses on digital twins, autonomous driving, intelligent transportation systems and human-autonomy teaming. He earned a doctorate in mechanical engineering from the University of California, Riverside, in 2019.