State estimate consistency monitoring in Gaussian filtering framework

The paper deals with the state estimation of the nonlinear stochastic dynamic systems by inherently approximate Gaussian filters. In particular, the stress is laid on the evaluation of the Gaussian filter state estimate consistency, which is a key indicator of the filter correct functionality and, thus, vital information for safety-critical applications. A novel on-line state estimate consistency monitoring test is proposed. Compared to the state-of-the-art tests, the proposed test directly works in the state-space domain without an assumption on the known true system state. Design of the test is free of specification or tuning of any threshold by the user. The proposed test is thoroughly analysed and compared with other tests on a theoretical and simulation basis. (C) 2018 Elsevier B.V. All rights reserved.