An efficient statistical adaptive order-switching methodology for kalman filters

Variants and improvements of the extended Kalman filter (EKF) always encounter a dilemma between the estimation accuracy delivered and the computational burden associated. Targeting to improve this shortcoming, achieving estimations with high accuracy and computational efficiency, this paper investigates a new adaptive high order extended Kalman filter (AHEKF) based on the design of a special order-switching strategy within one single filter run. At each filter step, an innovation-based function, accounting for the filter consistency, is put forward and estimated to determine the necessity of an order-switching operation. We think that the methodology could be of general applicability in procedures where a switching order approximation is possible, and in our work, simulations and performance evaluations are illustrated with a particular celestial mechanics problem dealing with orbit determination.