A Study on Acquisition of Behavior Patterns and Automation Proposing by Intelligent Agents Using IoT Devices

Abstract

The use of IoT devices is increasing, but some users are unable to fully utilize them. To support these users, we proposed a series of systems that collect data from multiple IoT devices, analyze the data to obtain behavior patterns, and automate device operations. To evaluate our proposed system, we conducted two evaluation experiments. In Experiment 1, we obtained participants′ behavior data during the seven-day period and acquired behavior patterns. In Experiment 2, we automated the device operations and measured the number of times the operations were automatically executed during the seven-day period. As a result, 294 behavior patterns were acquired in Experiment 1. We automated 6 of these patterns and these were automatically executed 501 times in Experiment 2. These results indicate that it is possible to acquire behavior patterns based on our proposed system and to automate device operations for acquired behavior patterns. There arises some issues with the validity of the analysis method, coordination among sensors, and partial manual operation. In the future, we aim to solve these issues and install Digital Twin to implement an intelligent agent that can acquire various behavior patterns with higher accuracy.