Design and experimental validation of an IoT-RPA-based smart refrigerator inventory system

Abstract

This paper presents the design and experimental validation of an integrated IoT-RPA cyber-physical system for automated food inventory monitoring and replenishment in a smart refrigerator. The proposed solution combines weight-based IoT sensing with robotic process automation to form a closed-loop "measurement-decision-action" architecture that minimizes user involvement. The hardware subsystem integrates a strain-gauge load cell, an HX711 analog-to-digital converter, and a WeMos D1 (ESP8266) microcontroller to measure product weight and transmit telemetry via Wi-Fi to a local Flask-based server. Measurement data are stored in CSV format and processed by a Robotic Process Automation (RPA) module implemented in Microsoft Power Automate Desktop, which initiates a digital ordering scenario when the product weight falls below a predefined threshold. Experimental validation under conditions close to real household operation demonstrated a mean absolute error below 2% and a coefficient of variation under 5%, as well as reliable transmission of more than 10,000 telemetry records without data loss. The RPA subsystem was evaluated through 120 execution cycles over seven days, achieving a success rate of 93.3% with an average execution time of 27.4 s under real web-interface constraints. The results confirm the feasibility, stability, and scalability of the proposed IoT-RPA architecture for autonomous household inventory management.