Skip to content

Latest commit

 

History

History
60 lines (48 loc) · 2.6 KB

README.md

File metadata and controls

60 lines (48 loc) · 2.6 KB
Raw

GP4AES - Gaussian Processes for Adaptive Environmental Sampling

This package implements an algorithm for Gaussian Regression-based Adaptive Sampling for environmental variables, in particular, those found in water bodies.

Installation

To install the package, make sure to have installed Python Poetry in your system. Then, run the following commands:

  1. Install the gp4aes library:
poetry install
  1. Run the demo script:
poetry run python scripts/DEMO.py datasets/finland_forecast_14_04_17_04.nc datasets/finland_ocean_color_14_04_17_04.nc config/finland_lres.json config/finland_hres.json results/LSQ_mission.h5 LSQ

Try each Submodule

Test each submodule individually.

Read .nc datasets into .h5 files and plot them

  1. Low Resolution data:
poetry run python scripts/read_nc_into_h5_file.py datasets/finland_forecast_14_04_17_04.nc config/finland_lres.json results/finland_lres.h5 1618610399
  1. High Resolution data:
poetry run python scripts/read_nc_into_h5_file.py datasets/finland_ocean_color_14_04_17_04.nc config/finland_hres.json results/finland_hres.h5 1618610399

Train kernel with data from previous days

Kernel is trained with LR data from the previous 3 days, and the parameters are printed in the terminal. Since operational area contains NaNs, it is necessary to choose a clipped area from the original, containing only ocean samples. Runs a multi-thread optimizer and can take up to 5/7 min:

poetry run python scripts/train_kernel_GP.py results/finland_lres.h5 1618610399 MAT 3 --clipped_area 61.412 61.80 20.771 21.31

A sample minimization iteration is the following parameters: 44.29588721 0.54654887 0.26656638

Prediction with scattered data

Predict HR data with Kernel trained as aforementioned. Measurements are scattered. To change datasets size, hardcode them on the script.

poetry run python scripts/gpr_interp.py results/finland_hres.h5 1618610399 MAT --kernel_params 53.24932783 0.48764177 0.2128522

Run script that plots ground truth, predicted mean, relative error and standard deviation:

poetry run python util/plot_gp.py results/finland_hres.h5

Remark: This step is not needed to run the trajectory script.

Compute trajectory

Compute trajectory based on the Gaussian Process estimation algorithm:

poetry run python scripts/run_mission.py datasets/finland_forecast_14_04_17_04.nc datasets/finland_ocean_color_14_04_17_04.nc config/finland_lres.json config/finland_hres.json results/GP_mission.h5 GP

To plot the results after the mission:

poetry run python scripts/plot_results_after_mission.py results/GP_mission.h5