| The overarching concept behind deploying sensor networks is to utilize capabilities provided by small-sized, low powered, and low cost battery operated sensor nodes that are capable of sensing, processing, and communication. With plenty of open challenges and outward-looking applications, wireless sensor networks (WSNs) have grown from a brand new discipline to play a central role in many scientific and technical fields of our society. Multiple small autonomous sensing devices, often called motes, can be used cooperatively and collaboratively to achieve a common goal in a highly efficient manner. These self-configuring motes usually have limited resources: computation, communication range, memory, and more importantly battery power. Thus, the nature of WSNs necessitates efficient strategy to design its applications, of which energy efficiency is paramount. Furthermore, the ad hoc and dense deployment of WSNs and the need for energy efficient strategy requires efficient organization of the network topology for the purpose of balancing the load and prolonging the network lifetime. Therefore, energy efficiency is a major design goal in these networks and node clustering is employed to meet such design goal.;Motivated by these concerns, considerable research efforts have been predominantly dealt with designing energy-aware techniques. Among those techniques, clustering has been proven to provide energy efficiency and long lifetime of the sensor networks. Specifically, WSNs are deployed in an ad hoc manner and have large number of sensor nodes. Hence, distributed clustering approaches that rely only on neighborhood information are preferred for sensor networks. Unfortunately, most of the existing clustering algorithm does not address the hot spot problem that arises in the vicinity of the base station. The nodes in the hot spot are required to relay disproportionately high traffic originated by the nodes that have no direct communication with the base station, and therefore depletes their energy quickly. This problem causes the network partition leaving an area of network uncovered.;In this thesis, we study and systematically evaluate the distributed clustering algorithm that employ hybrid of three parameters namely: residual energy, intra-cluster communication cost, and variable transmission power level to form clusters and to select the cluster head. We then proposed a solution to mitigate the hot spot problem. An unequal clustering mechanism is implemented among nodes at different distances to the base station, so that the hot spot problem can be mitigated by evenly distributing energy consumption in the sensor network. We conduct real world experiments to systematically evaluate the performance of our proposed model and their analysis is presented. |
