Large Language Models for Control and Decision Making ACC2025

About

LLM4Control Workshop

Welcome to 1^st LLM4Control Workshop co-located with ACC 2025!

LLM4Control workshop focuses on a timely and rapidly evolving intersection of artificial intelligence (AI) and control systems. Recent advancements in LLMs have significantly expanded the scope of what AI can accomplish in complex problem-solving, driving applications across domains such as coding, reasoning, mathematics, and science. As LLM capabilities continue to advance, their potential to transform control theory and practice is becoming increasingly evident and impactful. This workshop aims to provide a comprehensive review of various recently-developed results on the capabilities of LLMs in control engineering and decision making. By bringing together experts from control systems, natural language processing, and robotics, the workshop will foster interdisciplinary discussions that highlight how LLMs are being utilized to automate and enhance control design processes, optimize decision-making strategies, and address complex challenges across multiple domains. Participants will gain insights into LLM reasoning and applications for control engineering, as well as emerging opportunities for integrating LLMs into future control systems. Finally, the workshop will organize a panel to discuss the concerns on the accountability and reliability of LLM-based control design.

Specific topics covered in this workshop include but not limited to

capabilities of LLMs in answering textbook-level control system questions;
advanced techniques for LLM reasoning (e.g. Chain-of-Thoughts, Tree-of-Thoughts, Program-of-Thoughts, Reasoning-via-Planning, self-refinement, tool use, and LLM agents);
ControlAgent: combining LLM reasoning with control toolboxes for automating practical control design
combining LLM reasoning with verification tools for automating verification of control systems;
large agent models;
novel applications of LLMs and vision language models (VLMs) for robotic manipulations;
limitations of LLMs for control engineering;