An AI agent-enabled smart home control system with ESP32

Abstract

Smart home systems continue to suffer from limited cross-platform compatibility, insufficient personalization, and unreliable remote access, hindering the unified device management expected in modern living environments. This paper presents an AI agent-based home control system centered on the ESP32-S3 microcontroller. The system integrates environmental sensors, a voice interaction module, and actuators, and combines automatic speech recognition, a large language model for natural language understanding, and MQTT-based messaging to enable coordinated multi-terminal control. It supports three concurrent interaction modes—local voice commands, a mobile application, and third-party ecosystem integration—while providing environmental monitoring, conversational response, and remote management capabilities. A prototype implementation demonstrates low latency, reliable cloud-edge synchronization, and modular hardware extensibility. Owing to its low cost and flexible scalability, the system is suitable for residential, research, and educational deployments. These results indicate that integrating on-device AI inference with lightweight communication protocols offers a practical route toward interoperable and user-centric intelligent spaces.