MEKF_vs_MHE

Use simulated and clinical data to compare Multi Extended Kalman Filter and Moving Horizon Estimation to identify PD parameters in drug model during total intravenous anesthesia. This code is associated to the paper "Comparison of Multiple Kalman Filter and Moving Horizon Estimator for the Anesthesia Process" by Bob Aubouin--Pairault, Mirko Fiacchini, and Thao Dang. Please contact me for an early access to the paper or if you have any question. If you are using this code for research purpose, please cite the paper.

Structure

.
├── data                # Simulations data generated to compare the results
├── scripts             # python scripts to launch the tuning and test of the estimators 
├── observabilite       # scripts to study the observability of the system
├── realistic_simu_PID  # scripts to generate the simulated data 
├── LICENSE
├── requirements.txt
├── README.md
└── .gitignore     

Usage

First some folder must be created to store the data and the results.

mkdir data figures
mkdir data/mekf_p data/mhe_std data/simulations data/vital
mkdir data/vital/mekf_p data/vital/mhe_std

Code have runned using python 3.9.

Simulation data can be generated with the following command:

python3 realistic_simu_PID.py

Then the tuning of the hyper parameters can be launched with the following command:

python3 scripts/tuning_mekf.py
python3 scripts/tuning_mhe.py

The test of the estimators on simulated data can be launched with the following command:

python3 scripts/test_on_simu.py
python3 scripts/metrics_function.py

Results figures will be saved on the figures folder.

The test of the estimators on clinical data can be launched with the following command:

python3 scripts/download_vital.py
python3 scripts/test_on_vital.py
python3 scripts/metrics_vital.py

Results figures will be saved on the figures folder.

License

GNU General Public License 3.0

Project status

• Code finished, paper in review.

Author

Bob Aubouin--Paitault