Improved square root unscented Kalman filter and its application in rendezvous and docking
Guang‐Ren Duan · Dianji yu kongzhi xuebao · 2010
By integrating Gaussian process regression into the square-root unscented Kalman filter,a filter algorithm is deduced to resolve the problem that classical filter algorithms are restricted by the uncertainty of system model and noise covariance.In this algorithm,estimation stage adopting standard square-root unscented Kalman filter was composed of time update and state update,and the square-root of covariance matrix was propagated and updated by QR decomposition and Cholesky factor updating.State equation and observation equation were replaced by their regression models respectively,and corresponding noise covariance was adjusted adaptively.The simulation results show that the accuracy of the algorithm meets the request of navigation,and its effectiveness is verified.