A Fast Multidimensional Scaling Filter for Vehicular Cooperative Positioning

Abstract

Vehicular communication technologies are becoming staples of modern societies. This paper proposes a new positioning algorithm for vehicular networks. The algorithm is a non-classic Multi-Dimensional Scaling Filter (MDSF) that builds on a novel and computationally effective Multi-Dimensional Scaling (MDS) solution covariance estimation technique and also a Maximum Likelihood (ML) filter. In general a major drawback of the non-classic MDS is the high computational cost because of its iterative nature. It is shown that a special blend between vehicular Map-Matching (MM) and MDSF considerably reduces the number of iterations and the convergence time, making the MDSF a suitable algorithm for vehicular network positioning. The performance of MDSF is compared with that of an Extended Kalman Filter (EKF) together with the Cramar Rao Lower Bound (CRLB). It is shown through simulation that for all types of traffic conditions MDSF performs better than EKF and closer to CRLB than EKF. It is also shown that both MDSF and EKF algorithms are robust to typical Global Positioning System (GPS) outages in deep urban canyons. CRLB also proves that Cooperative Positioning (CP) in general has the ability to bridge short GPS outages.