Foundation Models for Control (FM4Control): Bridging Language, Vision, and Control

Foundation models (large language models, LLMs, and vision-language models, VLMs) now demonstrate an ability to understand, reason about, and generate artifacts tied to control systems. Leveraging them promises more natural human-machine interfaces, faster controller prototyping, and enhanced monitoring. Yet rigorous understanding of their capabilities, limitations, and safety implications remains scarce within the classical control community. This workshop gathers researchers from data-driven AI and control theory to chart a path that couples foundation-model expressiveness with principled modeling, estimation, and control.

Topics of interest include:

• Natural Language Interfaces for Control Specification

• Commonsense Constraints via Vision–Language Models

• FMs as Control Design Assistants

• Monitoring, Fault Detection and Diagnosis

• Generative Assets for Simulation and Testing

• Applications including robotics & manipulation, autonomous vehicles, traffic signal networks, smart–grid control, industrial manufacturing, healthcare automation