A Clustering Algorithm And Improvement Based On P-changed Affinity Propagation For Heterogeneous Wireless Sensor Networks

Wireless sensor networks are composed of wireless sensor nodes widely distributed in the environment, nodes communicate through wireless communication to format a multi-hop self-organizing network to perceive of objective physical world and transfer information. Wireless sensor networks are prone to fast deployment, strong applications, and good economic characteristics, having a broad prospects in the field of national defense, health care, environmental monitoring and so on, and has attracted widespread attention.Since the wireless sensor networks and the traditional wireless ad hoc networks are different, the requirement of routing protocol is different also. Due to the energy limitations of nodes, energy consumption has become the primary objective of the study. The routing protocol should not only design fine information transmission path, but also guarantee low energy consumption and node life balance.Clustering routing protocol is studied in this paper, and inspired by data clustering algorithm, affinity propagation clustering algorithm is applied to the wireless sensor network clustering, a new algorithm named CPAP (Clustering based on P-changed Affinity Propagation) is proposed. CPAP changes the conventional set mode of preference p, and considering both energy and distance while clustering. Besides, by analyzing and comparing the effects of the parameter K, its approximate optimal value is achieved. Simulation results show that compared with PECBA, which is also based on affinity propagation algorithm, the first node’s death time is delayed by28.5%in CPAP, which illustrates that the proposed solution makes a better utilization of energy before network is dead and enhances the energy efficiency.Affinity propagation algorithm is a centralized algorithm with a quadratic complexity, as the network size and the number of nodes increase, the algorithm complexity and operation cost increase, limiting the scope of application of the algorithm. Therefore, in chapter4, a distributed CPAP algorithm is proposed, the network is divided into grids, each grid uses CPAP algorithm to select the candidate cluster heads at the same time, and then the final cluster heads are selected from the candidates. The simulation shows that the distributed and centralized algorithms have the same clustering results, and distributed CPAP algorithm is feasible in a large-scale network. The energy consumption difference between two-level and multi-level network is compared and analyzed in this chapter. Finally, the energy consumption difference between single-hop and multi-hop is simulated and compared to put forward the best routing mode.