Exploiting the imperfect knowledge of reference nodes positions in range based positioning systems

In this paper, the problem of uncertainty on reference nodes positions is addressed in the context of hybrid data fusion techniques for localization. This problem arises in B3G networks where different location-dependent observables come from heterogeneous Radio Access Networks (RAN) leading to different levels of uncertainty on both ranges and anchor nodes positions. We assume a Gaussian model on the node position error as well as on the ranging error. We derive novel Maximum Likelihood based location estimator which considers these two sources of uncertainty. The performances of this new estimator is then compared to the ML estimator which does not consider erroneous reference nodes positions. Monte Carlo simulations show that the proposed estimator achieves better performances especially in the context of short range positioning.