Estimation of Time Varying Parameters Using Distributed Techniques

Abstract

Keeping in view, the increasing importance of wireless networks in modern application scenario, there is a increased necessity of improvement of error performances of these networks. To tackle this problem, many algorithms are developed. A possible solution is to use ADMM by considering problem statement as convex optimization. But due to sluggish convergence to high accuracy, Average Consensus algorithm is proposed,which gives better performance than ADMM. In this case,every sensor holds a local estimate of a global average, and update its value by performingga weightedgsum of the estimatesgof all itsgneighbors. The convergence speed of this algorithmsgdepends on thegweights used by every sensor.But the main problem arises in this algorithm is communication delay, which was overcome by using Gossip algorithm.In this algorithm, nognode communicates withgmore thangone neighbour ingevery time slot, whichgin turn reduces communication delays. Different applications which uses average consensus algorithm to solve problems are discussed in this work, which includes Distributed Consensus-Based Demodulation and Distributed PrincipalgSubspace Estimation ingWireless Sensor Networks. The experimental results shows these algorithms works well for solving problems including distributed demodulations.