Abstract:Recent research in nonlinear filtering and signal processing has suggested an efficient derivative-free Extended Kalman filter (EKF) designed for discrete-time stochastic systems. Such approach, however, has failed to address the estimation problem for continuous-discrete models. In this paper, we develop a novel continuous-discrete derivative-free EKF methodology by deriving the related moment differential equations (MDEs) and sample point differential equations (SPDEs). Additionally, we derive their Cholesky-based square-root MDEs and SPDEs and obtain several numerically stable derivative-free EKF methods. Finally, we propose the MATLAB-oriented implementations for all continuous-discrete derivative-free EKF algorithms derived. They are easy to implement because of the built-in fashion of the MATLAB numerical integrators utilized for solving either the MDEs or SPDEs in use, which are the ordinary differential equations (ODEs). More importantly, these are accurate derivative-free EKF implementations because any built-in MATLAB ODE solver includes the discretization error control that bounds the discretization error arisen and makes the implementation methods accurate. Besides, this is done in automatic way and no extra coding is required from users. The new filters are particularly effective for working with stochastic systems with highly nonlinear and/or nondifferentiable drift and observation functions, i.e. when the calculation of Jacobian matrices are either problematical or questionable. The performance of the novel filtering methods is demonstrated on several numerical tests.

What problem does this paper attempt to address?

This paper aims to solve the state estimation problem in continuous - discrete nonlinear stochastic systems. Specifically, when dealing with such systems, the traditional Extended Kalman Filter (EKF) needs to calculate the Jacobian matrices of the drift of the nonlinear model and the measurement vector function, which is a difficult problem in highly nonlinear and/or non - differentiable systems. To solve this problem, the author proposes a derivative - free Extended Kalman Filter (derivative - free EKF) method and derives the relevant Moment Differential Equations (MDEs) and Sample Point Differential Equations (SPDEs) for the continuous - discrete model. In addition, the author also derives the square - root MDEs and SPDEs based on Cholesky decomposition, thus obtaining a numerically stable derivative - free EKF method. Finally, the author provides MATLAB implementation methods of these algorithms, which use MATLAB's built - in numerical integrators to solve MDEs and SPDEs, ensuring the accuracy and adaptability of the implementation methods. ### Main contributions of the paper: 1. **Derivation of continuous - discrete derivative - free EKF**: Based on the principle of derivative - free EKF, the Moment Differential Equations (MDEs) and Sample Point Differential Equations (SPDEs) in the continuous - discrete model are derived, and the conventional MDE - and SPDE - based derivative - free EKF implementation techniques are obtained. 2. **Square - root versions of MDEs and SPDEs**: Further derive the square - root MDEs of Cholesky factor propagation, and several square - root implementation methods in the measurement update step, thus obtaining a numerically stable Cholesky - based square - root MDE - and SPDE - based derivative - free EKF method. 3. **MATLAB implementation**: Provide MATLAB implementation methods for all continuous - discrete derivative - free EKF algorithms. These methods use MATLAB's built - in numerical ODE solvers to ensure the accuracy and adaptability of the implementation methods, and users only need to set tolerance values without additional coding. ### Background and motivation of the paper: - **Problems of traditional EKF**: When dealing with nonlinear systems, the traditional EKF needs to calculate the Jacobian matrix, which is a difficult problem in highly nonlinear and/or non - differentiable systems. - **Advantages of derivative - free EKF**: The derivative - free EKF method approximates the probability description of state estimation by using a discrete vector set, avoiding the calculation of the Jacobian matrix, and is suitable for highly nonlinear and/or non - differentiable systems. - **Requirement for continuous - discrete models**: Existing derivative - free EKF methods are mainly for discrete - time systems, while continuous - discrete models are more common in many practical applications. Therefore, it is necessary to develop derivative - free EKF methods suitable for continuous - discrete models. ### Technical details of the paper: - **Moment Differential Equations (MDEs)**: Derive the MDEs for the prediction step, which describe the time evolution of the state mean and covariance matrix. - **Sample Point Differential Equations (SPDEs)**: Derive the SPDEs for the prediction step, which describe the time evolution of sample points and avoid directly dealing with the covariance matrix. - **Square - root implementation**: Through Cholesky decomposition, derive the square - root versions of MDEs and SPDEs to ensure numerical stability. ### Experimental results: - **Numerical tests**: The effectiveness and accuracy of the new method are verified through multiple numerical tests, especially when dealing with highly nonlinear and/or non - differentiable systems. In conclusion, this paper solves the limitations of traditional EKF in dealing with complex nonlinear systems by proposing the continuous - discrete derivative - free EKF method and its MATLAB implementation, providing new tools and methods for practical applications.

On derivative-free extended Kalman filtering and its Matlab-oriented square-root implementations for state estimation in continuous-discrete nonlinear stochastic systems

Continuous-discrete derivative-free extended Kalman filter based on Euler-Maruyama and Itô-Taylor discretizations: Conventional and square-root implementations

Continuous–discrete derivative-free extended Kalman filter based on Euler–Maruyama and Itô–Taylor discretizations: Conventional and square-root implementations

Continuous-discrete unscented Kalman filtering framework by MATLAB ODE solvers and square-root methods

MATLAB-based general approach for square-root extended-unscented and fifth-degree cubature Kalman filtering methods

Online chaotic time-series prediction with the derivative-free extended Kalman filter

Square-root filtering via covariance SVD factors in the accurate continuous-discrete extended-cubature Kalman filter

Constructing numerically stable Kalman filter-based algorithms for gradient-based adaptive filtering

SVD-Based Kalman Filter Derivative Computation

High-Order Extended Kalman Filter for State Estimation of Nonlinear Systems

State Estimation for Continuous-Discrete-Time Nonlinear Stochastic Systems

Iterated Kalman Methodology For Inverse Problems

Extended fractional singular kalman filter

Do the Contemporary Cubature and Unscented Kalman Filtering Methods Outperform Always the Traditional Extended Kalman Filter?

New version of continuous-discrete cubature Kalman filtering for nonlinear continuous-discrete systems

A Novel Case of Practical Exponential Observer Using Extended Kalman Filter.

Stirling's Interpolation Filtering and Extended Kalman Filtering for Orientation Estimation

Kalman Filter and its Modern Extensions for the Continuous-time Nonlinear Filtering Problem

A Novel Extended Kalman Filter for A Class of Nonlinear Systems

Distributed State Estimation Based on Unscented Kalman Filter

An Adaptive State Filtering Algorithm for Systems With Partially Known Dynamics