Consensus-based distributed unscented target tracking in wireless sensor networks with state-dependent noise

Most existing state estimation approaches assume that observation noise of sensors is independent on the state vector. However, in target tracking applications with ranging or bearing sensors, a more realistic approach is to consider the measurement noise to be state-dependent. In this paper, generalized unscented information filter (GUIF) is developed for target tracking in wireless sensor network (WSN). Nodes are assumed to be equipped with ranging and bearing sensors with their measurement noise to be dependent on sensor to target distance. To cope with state-dependent noise of sensors, the nonlinear observation model is proposed to be rewritten into a new multiplicative form. Using unscented transformation, the linearized form of the new observation model is derived. Then, new formulations of GUIF are derived for state estimation of the target. Next, the consensus technique is employed to derive a distributed implementation of GUIF. State estimation error of local estimators is then proved to be bounded in mean-square under network connectivity and collective observability assumptions. Effectiveness of the proposed estimator is also investigated by simulations.