Lee, Eun Kyung. Impact of neighborhood discovering and adaptive sampling in wireless sensor networks. Retrieved from https://doi.org/doi:10.7282/T36Q1XHG
DescriptionWireless Sensor Networks (WSNs) are networks characterized by a dense deployment of sensor nodes. Because of the dense deployment, sensors can make interference when exchanging data messages. Besides these data messages, in location-based routing that uses geographical positions to route messages, there is a Neighborhood Discovery Protocol (NDP). It should periodically broadcast "Hello" packets to discover neighboring nodes and maintain routing tables updated. This is due to the uncertainty of the wireless environment such as varying radio interference and mobility. Due to the overhead caused by these periodic broadcasts from many nodes in certain radio range, however, NDP may heavily impact on the performance of the routing scheme itself, which in turn could affect end-to-end performance. Although this is an important and challenging problem in WSNs, this impact and the associated tradeoffs have not been fully explored in the literature. Hence, in the first half of this thesis, an analytical and experimental study is conducted to determine how parameters such as power and transmission frequency of neighborhood discovery packets affect the communication process in static and mobile environments.
In addition, WSNs are used to monitor and reliably estimate a phenomenon from the collective information provided by its constituent sensor nodes. Due to the high density of the sensor nodes, the data obtained from them are usually correlated in both space and time. Adaptive sampling is a method that employs this spatio-temporal correlation inherent in WSNs to obtain an energy-efficient estimate of the field. In the second half of this thesis, a distributed, hierarchical, cluster-based adaptive sampling framework is proposed using multiple manifestations for field estimation in three-dimensional environment. Nodes sensing highly correlated values in space are grouped to form clusters and these clusters are modified based on variation in sensor data over time. Energy efficiency is achieved through minimization of communication costs by restricting data communication to the local domain (within clusters) and by applying sleep mode. Moreover, a phenomenon is more reliably captured by using multiple manifestations than by using a single manifestation. It ensures joint optimization by adaptively varying the sampling rates in both space and time domains.