Power Optimized and Power Constrained Randomized Gossip Approaches for Wireless Sensor Networks

Abstract

In this letter, we investigate the randomized gossip (RG) algorithm by taking the power consumption over wireless sensor networks (WSNs) into account for distributed averaging. The convergence of the classic RG problem can be improved by optimizing the second largest eigenvalue of the average update matrix, leading to the fastest distributed linear averaging (FDLA). As the total transmission power determines the lifetime of WSNs, we propose to jointly optimize the power consumption and the second largest eigenvalue, such that a trade-off between the power consumption and the convergence rate is obtained. Further, since each sensor node usually has a limited energy budget, we incorporate an additional constraint on the local power consumption for the FDLA formulation, such that the survival of nodes can be guaranteed. Numerical simulations using a WSN validate the effectiveness of the proposed methods.