Information-Intensive Control Method Using Estimated Observation Values Based on Correlation of Observation Values from Multiple Sensors

The authors have installed wireless terminals to monitor water levels in rivers and lakes in Shiojiri City, Nagano Prefecture, as a demonstration experiment of a wireless sensor network, and have been conducting environmental observations. Since the installation of the wireless sensor network, stable information aggregation has been confirmed. Wireless sensor networks for environmental observation require low power consumption and efficient use of wireless resources. Low power consumption is required because the environment in which they are installed has little power infrastructure, and the power supply is often powered by dry cell batteries or batteries. Efficient use of radio resources is also required because there may be other systems using the same frequency band. We have shown through previous data analysis that the observed information is correlated. In this presentation, we use this correlation to predict future observations by machine learning, and present a study of information aggregation methods that control the frequency of information aggregation to improve the efficiency of power consumption and wireless resource utilization.