Weighting strategy for MDS tracking in LoS conditions

In this paper we propose a low-complexity non-parametric pre-filtering/weighting technique for Time of Arrival (TOA)-based target tracking systems. The output of a wavelet transform computed on each one of the anchor-to-target links is used to estimate the noise and the dynamic for the corresponding link. Subsequently this information is used to weight and adaptively decide on the reliability of pre-filtered data. To take full advantage of the pre-filtering scheme proposed, this is used together with the optimization algorithm (SMACOF), a low-complexity solution to the multidimensional scaling (MDS) problem. Since both the pre-filtering scheme and SMACOF don´t rely on any a-priori knowledge, their combination results in a completely non parametric approach to the target tracking problem. The aforementioned technique is investigated and compared against a Bayesian state variable formulation of the problem implemented through an Extended Kalman Filter (EKF). The MDS-based tracking algorithm with the wavelet pre-filtering scheme here proposed, while remaining completely non-parametric is shown to overcome the EKF algorithm under different target dynamics and noise conditions.