Distributed Upper-Bound Information Filter with Generalized Unknown Disturbances in Sensor Networks

This paper presents the distributed state estimation problem in sensor networks with generalized unknown disturbances (GUDs) in measurement equations, where GUDs in different sensors are different and varied. For such problem, a centralized upper-bound filter is discussed via measurement augmentation by seeking a feasible solution of the scalar convex optimization with higher-dimensional matrices, which is further substituted by multiple scalar convex optimizations related to each sensor with lower-dimensional matrices under the criterion of trace operation. Meanwhile, instead of constructing the upper bound of estimate error covariance, the lower bound of the inverse matrix/covariance is pursued and the corresponding information filter form is presented under multiple sensors condition. Then, the distributed realization of the derived information filter is proposed via consensus algorithm to give a tradeoff between estimate accuracy and fast fusion. A target tracking example in sensor networks with different GUDs is simulated to testify the proposed method.