Adaptive Sampling in Wireless Sensor Networks for Air Monitoring System

Abstract: In wireless sensor networks (WSN), it is important to use resources efficiently because sensors have limited resources such as battery life and computational power. In this thesis, we study the method which can save energy of air-monitoring sensor networks with respect of QoS (quality of service). From historical data, we observe that during certain time of the day, concentration of air pollutants has no radical change, from which we can conclude that applying high sampling rate uniformly all the time is not necessarily required. Our approach uses Kalman filter technique to eliminate the noise from the sensor measurements, and adjust the sampling interval based on the difference between the present and previous measurements. If the sampling interval is within the sampling interval range, we use the new sampling interval for the next measurement and if not, a central server assigns a new sampling interval and sampling interval range to the requesting sensor. This way, we can achieve adaptive sampling based on input characteristics so as to save energy of the sensor net- work and also to obtain proper accuracy of sensor measurements. We simulated our method with real measurement data with Matlab and finally implemented our method in the GreenIoT project to demonstrate the energy- efficiency and sensing-quality of our technique.