A Set-Membership Filter with Box Bounding for Nonlinear Biased Dynamic Systems

This paper investigates a set-membership filter with box bounding for some nonlinear dynamic systems with sensor biases. This tracking problem is difficult since the uncertainty of both noises and biases. First, the sensor bias is augmented as part of the state. It is proved that when the parameter is bounded in a box, the remainder bound of a typical measurement function can be obtained analytically in real time. Next, both the state prediction and set-membership filter update can be equivalently transformed to solve linear optimization problems with lower computational complexity. Finally, a numerical example in the context of multiple sensor tracking system with sensor biases demonstrates the effectiveness of the box bounding state filter.