Distributed Recursive Filtering Over Sensor Networks With Nonlogarithmic Sensor Resolution

Sensor resolution, which is one of the most important parameters/specifications for almost all kinds of sensors, plays an important role in any signal processing problems. This article deals with the distributed filtering problem for a class of discrete time-varying stochastic systems subject to nonlogarithmic sensor resolution and stochastic nonlinearities. The soft measurement technique is exploited in the filter design to overcome the difficulties resulting from the sensor-resolution-induced (SRI) uncertainty. The aim of the presented filtering problem is to construct the distributed filter over a sensor network such that in the presence of SRI uncertainty and stochastic nonlinearity, an upper bound on the filtering error covariance is guaranteed and subsequently minimized by appropriately designing the filer parameters at each time instant. Moreover, a matrix simplification method is utilized to tackle the difficulties stemming from the sparsity of sensor networks. Finally, a numerical example is employed to illustrate the effectiveness of the proposed filtering scheme.