A Semidefinite Relaxation Method for Localization in Vehicular Ad-Hoc Networks

Abstract

Accurate localization is one of the most important premises for the application of vehicular Ad-Hoc networks. This paper explores the information fusion of the GPS measurements and the Radar TOA measurements to improve the localization accuracy. However, the formulation based on the Maximum-Likelihood method is a non-convex optimization problem. We reformulate it as a homogeneous quadratic programming with homogeneous quadratic equality constraints, which is non-convex and NP-hard. Therefore we propose a semidefinite relaxation method which can efficiently solve this non-convex problem. The Monte Carlo simulations are carried out and the results demonstrate that the proposed SDR method can achieve the CRLB accuracy.