Heterogeneous distributed clustering in sensor networks

Many clustering algorithms in various contexts have been proposed for wireless sensor networks but they usually assume that the environment is homogeneous, i.e. only specific and related data are transmitted. Furthermore, these approaches only consider a particular clustering parameter e.g. cluster diameter, load and spatial positions. As wireless sensor networks are widely extending in terms of broadness, heterogeneity, multitasking, and etc., single-purpose clustering seems not to be effective. Moreover, this is of more interest to form clusters based on various parameters and desires dynamically. In this paper, we propose a general network clustering approach which is able to operate in heterogeneous environments and clusters network nodes by considering multi-task and multi-parameter characteristics. The proposed solution takes some desired clustering parameters as its input and fuzzily decides how to form clusters of the network nodes. We formalize this solution and introduce a problem-unrelated distributed algorithm. This is also proved that in a very dynamic and noisy environment, the performance of the algorithm degrades at most by the factor of p where p is the probability of accessing a neighboring node.