This project presents the code and tests results performed for the submission of the paper : Integration of Large Language Models into Control Systems for Shared Appliances. The study aims at providing a solution to pilot various appliances with LLMs involved in the control loop. As a case study, we focused on the control of a heating system shared among multiple users.

Two case studies are presented :

1. the first focuses on the simulation of a heating system controlled by three personas,
2. the second focuses on mediation of the interaction user(s) and the heating system using ChatGPT.

This repository contains all archives related to the ICPR paper submission. The structure of this git is as follows:

• The supplementary-material.pdf contains the supplementary material for the paper,
• The case-study-1 directory contains all raw results about the heating simulation case study,
• The case-study-2 directory contains all raw results about the mediation w/ ChatGPT case study,
• The llm-cps-ac directory contains the main source code for the project.

This study has been conducted by Frédéric Montet and Karl Löwenmark with equal contributions and a fruitful exchange of knowledge between HEIA-FR (Fribourg, Switzerland) and LTU (Luleå, Sweden). This work has been supervised by Marcus Liwicki, Fredrik Sandin and Jean Hennebert.

A warm thank you to the HEIA-FR and LTU teams for their support and guidance throughout this project. As well as to the student and professors who supervised the OptiBot project. Also, a special thank you to the yord.ch team for their collaboration and support.

• Student: Philippe Marziale - [email protected]
• Supervisor: Houda Chabbi - [email protected]
• Supervisor: Sébastien Rumley - [email protected]
• yord: yord.ch