The Impact of Clustering in Distributed Topology Control

摘要

Topology control is the problem of assigning power levels to the nodes of an ad hoc network so as to maintain a specified network topology while minimizing energy consumption (either minimizing the maximum power used by any node or minimizing the total (i.e. average) power used by the nodes). In [18], a hybrid framework for distributed topology control based on clustering was proposed. That framework, called CLTC, specifies algorithms for both 1-connected and 2-connected topologies, and works with any clustering algorithm. CLTC utilizes centralized topology control within each cluster, but is otherwise fully distributed, hence the characterization of the method as hybrid. This paper studies the effect of six representative clustering methods on the quality of the topology control solutions provided by CLTC. The results establish that the most important factors in determining the performance of CLTC are the average cluster size and the closeness of nodes in clusters. This leads to a tradeoff between the energy consumption, the complexity of cluster formation, and the scope to which the operations of CLTC are fully distributed. The paper also shows that, in general, there is a considerable increase in power usage (in the vicinity of 150%) by requiring a 2-connected network versus a 1-connected network.