Comparative Study Of Clustering Protocols With Probabilistic Model Checking

Wireless sensor networks with hundreds or thousands of sensor nodes have emerged as important computing platforms in recent years which can be applied in a wide range of areas. A wireless sensor network is grouped into several, usually disjoint clusters. Clustering techniques aim to support scalability and increase network lifetime for wireless sensor networks. Each cluster is managed by a cluster head which is elected by the sensor nodes based preferably on local information. Cluster heads are obliged for coordination among the nodes in their clusters and representation of their clusters to communicate with external observers. A large number of clustering protocols in wireless sensor networks are available currently.In this thesis, we analytically evaluate and compare four state-of-the-art of clustering protocols in wireless sensor networks, i.e., LEACH, GEN-LEACH, HEED and PANEL. Based on our own analytical framework that consists of a network configuration and an energy consumption model, we aim to analyze the four clustering protocols’functionality and performance.Our comparative analysis proved some properties of the investigated protocols that were foreseen based on previous simulation results. Such a proof is delivered for the expectation that the distribution of initial energy (whether it is the same or different for the nodes) indeed strongly influences the energy consumption of HEED and PANEL, or that the topology does not influence the number of cluster heads elected in case of LEACH, GEN-LEACH and PANEL. Moreover, we list such main factors influencing the behaviors of HEED.Two directions need work in the future:one is to explore more techniques such as model abstraction to deal with the huge state space for the clustering protocols with complicated logic, the other is to expand the comparative study to more clustering protocols after a systematic classification.