Marginal Distributionally Robust Kalman Filter (MDRKF)

This repository contains the implementation of the Marginal Distributionally Robust Kalman Filter (MDRKF), a robust estimation technique for systems with uncertain parameters. The MDRKF algorithm combines the principles of Kalman Filtering with Distributionally Robust Optimization to provide enhanced estimation performance under uncertainty.

Installation

To get started with the MDRKF project, clone this repository to your local machine:

git clone https://github.com/lylechan42/MDRKF.git
cd MDRKF

Usage

The primary scripts for the MDRKF algorithm are:

• dynamic_estimation_MC.m: Implements the Monte Carlo Simulation to evaluate our methods for dynamic estimation.
• dynamic_estimation_test.m: Test script to run our MDRKF algorithm briefly.

Utilities

The utils directory contains various utility functions used in the MDRKF algorithm:

• DRO.m: Distributionally Robust Optimization implementation.
• Frank_Wolfe.m: Frank-Wolfe algorithm for convex optimization.
• KF.m: Standard Kalman Filter implementation.
• MDRO.m: Marginal Distributionally Robust Optimization implementation.
• genSigma.m: Function to generate covariance matrices.
• getCorrData.m: Function to retrieve trajectory with correlated measurement.
• getData.m: Function to retrieve trajectory with uncorrelated measurement.
• getRanData.m: Function to retrieve trajectory with random parameters.
• getSys.m: Function to retrieve system parameters.
• isPSD.m: Function to check if a matrix is positive semi-definite.
• my_dist.m: Custom distance function.
• reconstruct.m: Function to reconstruct estimation matrices.
• save_estimations.m: Function to save estimation results.
• tau_update.m: Function to update estimation with tau divergence based DRO.

License

This project is licensed under the MIT License - see the LICENSE file for details.