Adaptive Generalized Robust Iterative Cubature Kalman filter

In this paper, an Adaptive Generalized Robust Iterative Cubature Kalman filter (AGRICKF) is developed to solve the problem of non-Gaussian noise and nonlinear problem employed in state estimation. The general robust loss function (GRL) can adjust the shape of the kernel in real time according to the noise environment, but its performance is still degraded when facing more complex noise environments such as Laplace noise. Therefore, a generalized GRL is further constructed by adding a shape modifier to further modify the shape of the kernel according to the noise environment, so as to effectively improve the ability of the algorithm to face complex noise environments. And the accuracy of the algorithm in nonlinear state estimation problems is enhanced through constructing nonlinear augmented model ensemble prediction error and measurement error. The experimental results confirm that the algorithm proposed in this paper has stronger robustness.