Performance evaluation of a novel adaptive variable structure state estimator

Abstract

An improved nonlinear adaptive state estimator called Adaptive Smooth Variable Structure Filter ASVSF has been proposed and its algorithm described. ASVSF extends the functionality and performance of a previously reported robust smooth variable structure filter (SVSF) with optimal boundary layer (SVSF-OBL). Improvement in performance includes the provision for accommodating unknown and time varying process noise covariance, which generally characterizes modelling uncertainty. The novelty of this proposed ASVSF estimator, which inherits the features of the SVSF, is that it adaptively provides an estimate of the unknown time varying process noise covariance (and hence called adaptive SVSF or ASVSF) which is required for determining the optimal boundary layer width of SVSF-OBL thus obviating the need of the prior knowledge of the process noise covariance. This makes the proposed estimator performance to be insensitive to (and therefore robust with respect to) unknown time varying process noise covariance while retaining the optimality of SVSF-OBL. The performance of the proposed ASVSF estimator is evaluated using Monte Carlo simulation and is compared with previously reported state estimators using a case of maneuvering civilian aircraft where a simplified and grossly approximate process model is used in the estimator/filter for tracking and thereby generating a time varying and unknown process noise covariance situation. Three different measures of Root Mean Square (RMS) error over the trajectory have been used for comparison which demonstrates the strengths of the proposed ASVSF estimator.