Low Power Wireless Sensor Network

Wireless sensor networks (WSN) take on an invaluable technology in many applications. Their prevalence, however, is threatened by a number of technical difficulties. In particular, the shortage of energy in sensors is a serious problem to which many solutions have been proposed in recent years. This thesis takes this area of research one step further and proposes solutions to better conserve energy in sensors. The research conducted can be divided into two parts. The first part is on the design and development of low-power sensors and communication devices capable of monitoring the environment. In this part of research, we first show how smartphones can be employed as a device to acquire data from low-power sensors. Then, by using the idea of duty cycling, we achieve a significant reduction in power consumption in environmental sensing. The second part of this research is on the use of data-driven approaches where scholars suggest reducing the amount of required communication so that more energy can be saved in sensors. The main idea is that the components of a sensor, including its radio, can be turned off most of the time without noticeable influence on the judgments made using the sensed data. In fact, the data not sensed when the sensor is powered down can be predicted using the computational intelligence methods. To do so, we employ a multi-layer perceptron to predict missing environmental data on the basis of what is sensed. We also show that the effectiveness of this technique highly relies on the correlation between the points making the time series of sensed data. Our experimental results evidence the usefulness of the technique we propose in the second part of this research. Indeed, we train a nonlinear autoregressive network against various datasets of sensed humidity and temperature in different environments. It is then observed that sensors can be powered on intermittently without any significant influence on the desired behavior of the sensor network. By testing on actual data, it is shown that the predictions by the device greatly obviates the need for sensed data during sensors' idle periods and saves over 65 percent of energy. It is also established that, among the solutions already proposed, the data- driven approach is best suited to Wireless Sensor Networks especially environmental sensin