Development of Monitoring System Based on IoT Devices Equipping with Heterogeneous Sensors for Robot Conversation According to Daily Activities of an Older Adult Living at Home

Abstract

We have been developing a home monitoring system in which a robot speaks to an older adult when anomaly is detected. In this paper, new stationary IoT sensor nodes were developed. The sensor nodes were powered by commercial electricity and equipped with privacy-friendly heterogeneous sensors measuring power consumption of electrical appliances, vibration, sound level, illuminance, pressure, temperature, humidity, and CO₂. We also programmed an IoT gateway to collect data every second from multiple nodes via Wi-Fi and transfer it to a database. Operational verification experiments revealed accurate measurements and uninterrupted operation for one month with data loss below 0.05% and within one minute. An anomaly detection experiment using supervised machine learnings demonstrated ability to detect various simulated anomalies embedded into accumulated data, achieving a recall rate exceeding 0.95. These results indicated the potential for the developed system to serve as a foundation for the targeted monitoring system.