GaFT Toolbox
Description
GaFT is an advanced MATLAB® toolbox that provides a numerically stable implementation of basic and advanced estimation algorithms accompanied with supporting tools for easy implementation and evaluation of estimation experiments.
It is designed and developed owing to long term experience in the design of advanced estimation methods and their application in a number of projects. This toolbox finds its use in the estimation task for user-defined models possibly using real data, assessing the suitability of different estimation methods for a given model, and easy development of new estimation methods.
Toolbox capabilities
- natural description of the problem
- processing the real data and generating simulated data
- straightforward experiment setup
- evaluating the quality of the estimate based on different measures
- assessing the suitability of the chosen estimation methods
Target application area
In the areas where it is necessary to estimate unknown variables such as in tasks requiring estimates of vehicle position, velocity and orientation in space. The implementation of these methods is useful in the application areas of navigation, process control, signal processing and prediction.
GaFT highlights
- versatility and ease of use
- implementation of popular estimation methods (EKF, UKF)
- emphasis on numerically stable implementation
- simple and straightforward experiment setup
- simple extensibility with custom estimation algorithms
- many examples presenting estimation experiment setup
The benefits of the toolbox
The toolbox was developed with emphasis on the quality of the estimation algorithms implementation and ease of use of the tool. It provides a proven numerically stable implementation of the individual estimation algorithms. Each aspect of the problem is covered by a corresponding component allowing natural description of the problem. Despite the versatility of the problem description the computational complexity of estimation experiments is not significantly affected.
GaFT components
The toolbox is written using object oriented methodology and the various classes can be grouped into following components.
The first three components at the top of the diagram serve for the description of the problem. These components make it possible to describe vast variety of problems ranging from time invariant linear systems to time varying nonlinear system with non-additive noises. The core elements are function classes used for description of the system structure and/or of the parameters of probability density functions.
The toolbox offers the implementation of the basic estimation methods and their numerically stable implementation in the Estimators component. The core class makes further extension of the toolbox with new methods easy.
The last component facilitates the execution of Monte Carlo simulation and evaluation of the results. The key part of this component is the implementation of various performance indices.
Implemented local estimation methods
Currently, the GaFT toolbox offers the following implementations of local estimation methods:| Estimator | Toolbox object | Note |
|---|---|---|
| (extended) Kalman filter | gaftKalman | Joseph form implementation of filter and RTS smoother |
| Unscented Kalman filter | gaftUKF | standard version of UKF with RTS smoother |
| (extended) square-root Kalman filters | gaftSRKalman | Bierman-Thorton and two Carlson-Schmidt implementations with deccorelation of measurements |
| square-root unscented Kalman filter | gaftSRUKF | UKF with square-root algorithm without decorrelation of measurements |
