Active consensus over sensor networks via selective communication

Distributed average consensus has been widely studied for sensor network applications. Whereas the asymptotic convergence rate has been extensively studied in prior work, other important and practical issues, including energy efficiency, have received relatively less attention. This paper presents a scheme for distributed consensus that can achieve a good balance between convergence rate and energy efficiency. The approach selects a subset of links with significant contribution to the formation of consensus at each iteration, thus adapting the network topology dynamically to the changes of the sensor states. We formulate a global optimization problem for optimal link selection, and subsequently factorize it into sub-problems that can be solved locally, via approximation. We derive an algorithm that can solve them efficiently by quadratic programming (QP) relaxation and random sampling. Simulations on both uniform-degree and nonuniform-degree random networks demonstrate that the proposed method can significantly reduce the communication energy cost, while maintaining a reasonably high convergence rate.