Adaptive Kalman Filter for Navigation Sensor Fusion

摘要: As a form of optimal estimator characterized by recursive evaluation, the Kalman filter (KF) (Bar-Shalom, et al, 2001; Brown and Hwang, 1997, Gelb, 1974; Grewal & Andrews, 2001) has been shown to be that minimizes variance estimation mean square error (MSE) widely applied navigation sensor fusion. Nevertheless, in designs, divergence due modeling errors is critical. Utilization KF requires all plant dynamics noise processes are completely known, process zero white noise. If input data does not reflect real model, estimates may reliable. The case theoretical behavior its actual do agree lead problems. For example, if provided with information behaves certain way, whereas, fact, it different will continually intend fit an incorrect signal. Furthermore, when measurement situation provide sufficient estimate state variables system, other words, covariance matrix becomes unrealistically small disregards measurement. In various circumstances where there uncertainties system model description, assumptions on statistics disturbances violated since number practical situations, availability precisely known unrealistic fact modelling step, some phenomena disregarded way take them into account consider nominal affected uncertainty. highly depends predefined models forms major drawback. agree, problems tend occur. adaptive algorithm one approaches prevent problem precise knowledge available. To fulfil requirement achieving optimality or preventing filter, so-called (AKF) approach (Ding, 4