Networked Strong Tracking Filtering with Multiple Packet Dropouts: Algorithms and Applications

This paper focuses on the design problem of a recursive networked strong tracking filter (NSTF) for a class of nonlinear networked systems with parameter perturbations and unknown inputs. The sensors for the networked system are allowed to be spatially distributed in a large geographical area, and signals are transmitted via a shared communication channel with limited capacity. For this kind of system structure, the measurements from different sensors may experience probabilistic data loss with different probabilities. A series of Bernoulli sequences is employed to describe the multiple packet dropout rates. Parameter perturbations and unknown inputs in the system are considered in the filter design process. A recursive networked extended Kalman filter is first derived in the least mean square sense by taking the packet dropout phenomenon into account. Then, a fading factor is introduced in the filter structure in order to cope with the parameter perturbations and unknown system inputs, and a recursive NSTF is derived by developing the so-called networked orthogonal principle. It is shown that the proposed NSTF is capable of providing satisfactory estimation results even in the presence of system parameter perturbations and/or unknown system inputs. A simulation study is carried out on a practical Internet-based three-tank system, and the estimation results show the effectiveness and applicability of the proposed filtering techniques.