Cooperative Localization in WSNs: a Hybrid Convex/non-Convex Solution

TL;DR

This paper introduces a two-stage distributed localization algorithm for wireless sensor networks that combines convex relaxation with a non-convex refinement, improving speed and accuracy.

Contribution

It presents a novel hybrid convex/non-convex approach for peer-to-peer localization that is fully distributed and enhances convergence speed and localization accuracy.

Findings

Significant speed improvement over previous methods

Effective localization accuracy demonstrated in tests

Robustness to network variations

Abstract

We propose an efficient solution to peer-to-peer localization in a wireless sensor network which works in two stages. At the first stage the optimization problem is relaxed into a convex problem, given in the form recently proposed by Soares, Xavier, and Gomes. The convex problem is efficiently solved in a distributed way by an ADMM approach, which provides a significant improvement in speed with respect to the original solution. In the second stage, a soft transition to the original, non-convex, non relaxed formulation is applied in such a way to force the solution towards a local minimum. The algorithm is built in such a way to be fully distributed, and it is tested in meaningful situations, showing its effectiveness in localization accuracy and speed of convergence, as well as its inner robustness.