CONTROL OF ELECTROTECHNICAL DEVICES BY LARGE LANGUAGE MODELS (LLM)
Oleksandr Shtykalo, Iuliia Yamnenko · Transactions of Kremenchuk Mykhailo Ostrohradskyi National University · 2024
This article discusses how to integrate the ESP32 microcontroller with large language models (LLMs) such as the GPT-3.5-turbo to automate tasks in IoT systems.The ESP32's capabilities in device management and communication via Wi-Fi and Bluetooth make it suitable for such integration.The article discusses the components involved, including the features of the ESP32 microcontrollers and the LLM architecture.The purpose of the study is to integrate the ESP32 with the GPT-3.5turbo on the example of intelligent control of a light sensor system.In today's IoT world, the ESP32 is gaining popularity due to its built-in wireless connectivity, and combining it with the LLM can increase the level of automation and task optimization.An experiment was conducted where a light sensor system based on ESP32 is controlled by instructions processed by the GPT model.The methodology involved the use of the ESP32-WROOM microcontroller programmed in the MicroPython environment.To set up the GPT-3.5-turbo,an instruction was written to control the lighting.The circuit included red and blue LEDs, a photodiode, and an ESP32 that was connected to Wi-Fi and communicated with the GPT-3.5-turbo.The results of the experiment showed successful control of the light sensor system using LLM.When the illumination increased, the system received a "high" response from the GPT-3.5-turbo, the blue LED went out and the red LED came on.When the illumination decreased, the system received a "low" response, which led to the reverse change of LEDs.All stages of the test showed compliance with the expected results.The originality of this study lies in the use of neural networks to control microcontroller systems.This opens up new opportunities for intelligent IoT solutions, where microcontrollers are able to perform complex tasks without significant dependence on data processing servers.The practical value of the work lies in the ability to automate home IoT systems and optimize tasks.The use of LLM for lighting control demonstrates the potential for creating intelligent home assistants.The findings confirm the success of controlling the light sensor system based on ESP32, which interacts with GPT-3.5-turbovia Wi-Fi.A script for the microcontroller and instructions for the GPT model have been developed for the system to function.Further research will focus on more complex systems with many types of data and investigate the use of fuzzy logic with LLM in IoT automation.