Energy Efficient Consensus Over Complex Networks

The need to extract large amounts of information from the environment to have precise situation awareness and then react appropriately to certain events has led to the emergence of complex and heterogeneous sensor networks. In this context, where the sensor nodes are usually powered by batteries, the design of new methods to make inference processes efficient in terms of energy consumption is necessary. One of these processes, which is present in many distributed tasks performed by these complex networks, is the consensus process. This is the basis for certain tracking algorithms in monitoring and control applications. To improve the energy efficiency of this process, in this paper we propose a new methodology to optimize the network topology. More specifically, the topologies we obtain are Pareto-optimal solutions in terms of energy consumption and network lifetime metrics. This methodology is first approached from a general point of view, including most network properties at a time. Then, since in the practice not all networks present the same characteristics, we identify three real settings in which the optimization must be tackled differently. This leads to three particularizations of the problem, where the appearance of well-known graph models: small world, scale free and random geometric graphs is related with certain environment and nodes characteristics. Finally, extensive numerical results are presented to show the validity and efficiency of the proposed methodology.