A Novel Set-Member Filter for Maneuver Non-Ellipsoid Extended Target Tracking With Unknown but Bounded Noises

For non-ellipsoid extended target tracking, the process disturbances and measurement noises are described by independent random variables, i.e., Gaussian distribution. In many real scenarios, however, the disturbances and noises are completely unknown except for bounds. This article proposes a novel set-member filter for maneuver non-ellipsoid extended target tracking with unknown but bounded (UBB) noise. The non-ellipsoid extent of the extended target is described by the random hypersurface model (RHM), the predicted state and shape parameters are obtained by calculate the Minkowski sum of the ellipsoids of the state and shape parameters and UBB noises, respectively. At the measurement update stage, the predicted shape parameters are modified based on the maneuver angle, and the minimal ellipsoid containing the pseudo-measurement nonlinear function is calculated by solving the semi-positive definite programming problem, in which the trace of the shape matrix of the pseudo-measurement nonlinear function is as the objective function. The update shape parameters are updated by calculating the intersection set of predicted shape ellipsoid and the minimal pseudo-measurement ellipsoid. An evaluation of performance in a simulated environment shows that the proposed SMF-NEET approach outperforms the existing adaptive extended Kalman filter based on a RHM (RHM-AEKF) and unscented Kalman filter based on the random hypersurface model (RHM-UKF) methods with up to 83.9% and 88.9% improvement in the accuracy in position, 40.3% and 58.2% in velocity, 26.6% and 31.5% in the shape, respectively, while tracking a maneuver non-ellipsoid extend target with UBB Noises.