13 June 2023 in IEEE Signal Processing Letters
In this paper, we develop OR-IEKF which is a novel outlier-robust iterative extended Kalman filtering (IEKF) framework based on nonlinear regression formulation of update step. A new Kalman-type update step with reweighted prediction covariance and reweighted observation noise covariance are produced under the OR-IEKF framework, which could cut off the large outliers in observations causing by unknown outlier noises. By using various robust cost functions to solve such special nonlinear regression problems, we derive three algorithms. The performances of these new filters are evaluated in a nonlinear system simulation study.
16 February 2022 in IEEE Transactions on Neural Networks and Learning Systems
In this article, we investigate the approximation ability of recurrent neural networks (RNNs) with stochastic inputs in state space model form. More explicitly, we prove that open dynamical systems with stochastic inputs can be well-approximated by a special class of RNNs under some natural assumptions, and the asymptotic approximation error has also been delicately analyzed as time goes to infinity. In addition, as an important application of this result, we construct an RNN-based filter and prove that it can well-approximate finite dimensional filters which include Kalman filter (KF) and Beneš filter as special cases. The efficiency of RNN-based filter has also been verified by two numerical experiments compared with optimal KF.
12 April 2022 in IEEE Transactions on Aerospace and Electronic Systems
In this paper, we derive an optimal transportation particle filter for linear time-varying systems with correlated noises. This method can be regarded as the extension of the feedback particle filter with optimal transportation structure. However, the particles in our method are evolved in a deterministic way while we need to generate random particles in feedback particle filter. Consequently, we only need a very few particles to obtain the satisfying results and this property is especially significant for high dimensional problems. The error analysis of our method and feedback particle filter have been carried out when the system is time-invariant. Compared with feedback particle filter and ensemble Kalman filter, our method shows great efficiency in numerical experiments including both scalar case and high dimensional case.
10 January 2023 in IEEE Transactions on Neural Networks and Learning Systems
In this article, we propose the novel neural stochastic differential equations (SDEs) driven by noisy sequential observations called neural projection filter (NPF) under the continuous state-space models (SSMs) framework. The contributions of this work are both theoretical and algorithmic. On the one hand, we investigate the approximation capacity of the NPF, i.e., the universal approximation theorem for NPF. More explicitly, under some natural assumptions, we prove that the solution of the SDE driven by the semimartingale can be well approximated by the solution of the NPF. In particular, the explicit estimation bound is given. On the other hand, as an important application of this result, we develop a novel data-driven filter based on NPF. Also, under certain condition, we prove the algorithm convergence; i.e., the dynamics of NPF converges to the target dynamics. At last, we systematically compare the NPF with the existing filters. We verify the convergence theorem in linear case and experimentally demonstrate that the NPF outperforms existing filters in nonlinear case with robustness and efficiency. Furthermore, NPF could handle high-dimensional systems in real-time manner, even for the $100$ -D cubic sensor, while the state-of-the-art (SOTA) filter fails to do it.
8 July 2024 in IEEE Transactions on Automatic Control
In numerous application areas, high-dimensional nonlinear filtering is still a challenging problem. The introduction of deep learning and neural networks has improved the efficiency of classical algorithms and they perform well in many practical tasks. However, a theoretical interpretation of their feasibility is still lacking. In this paper, we exploit the representational ability of recurrent neural networks and provide a computationally efficient and optimal framework for nonlinear filter design based on the Yau-Yau algorithm and recurrent neural networks. Theoretically, it can be proved that the size of the neural network required in this algorithm increases only polynomially rather than exponentially with dimension, which implies that the Yau-Yau algorithm based on recurrent neural networks has the ability to overcome the curse of dimensionality. Numerical results also show that our method is more competitive than classical algorithms for high-dimensional problems.
23 October 2024 in IEEE Transactions on Automatic Control
This note introduces the Modified Extended Kalman Filter (MEKF), reformulating the EKF update step within a nonlinear regression framework. We propose a novel outlier-robust scheme, MCIC-MEKF, utilizing the minimum correntropy-induced cost criterion. We provide a theoretical analysis of its outlier robustness through stochastic stability, proving exponentially bounded mean square posterior estimation error under natural conditions. Additionally, we present a technical approximation for the adaptive Kalman gain, enhancing efficiency without compromising performance. Simulation results confirm MCIC-MEKF's robustness against various non-Gaussian noises with large outliers, outperforming several filtering benchmarks.