“AI ASSISTED PERSONAL ROBOT”

Abstract: The rapid growth of intelligent personal assistants has largely been driven by cloud-based artificial intelligence services, which inherently suffer from privacy concerns, network dependency, and non-deterministic latency. These limitations become critical in mobile robotic platforms where timely and reliable responses are essential for safe navigation and natural interaction. This paper presents the design and implementation of an AI-assisted personal robot that integrates a locally hosted Large Language Model (LLM), specifically Llama 3, with an ESP32-S3 microcontroller performing edge computing. The proposed system employs an I2S-based digital audio acquisition pipeline using an INMP441 microphone, a WebSocket-based bidirectional communication framework between the robot and a local AI server, and an ultrasonic-based safety interrupt mechanism for collision avoidance. Speech is captured as raw PCM data on the ESP32-S3, transmitted over Wi-Fi to a local workstation running Whisper for speech-to-text conversion and Llama 3 for intent understanding, and returned as structured JSON commands for real- time motor control. Experimental evaluation demonstrates sub-second end-to-end latency, robust command recognition in the presence of motor noise, and reliable indoor navigation under constrained network conditions. The results indicate that local LLM-driven robotics on commodity hardware can achieve low-latency, privacy-preserving interaction without dependence on cloud infrastructure, providing a viable architectural template for future embodied AI systems

Keywords: ESP32-S3, Edge AI, Local LLM, I2S Protocol, Web Socket Communication, Privacy-Preserving Robotics