Automatic-Data-Processing-for-Space-Robotics-Machine-Learning

Welcome to the repo for the paper titled "Automatic Data Processing for Space Robotics Machine Learning"! The paper can be found here: https://arxiv.org/abs/2310.01932

This repo contains PyQGIS code for generating viewsheds of the Curiosity or Perseverance rovers' points of view from a given NASA PDS image and label.

conda setup

$ conda create --name automatic-data-processing python=3.10.12
$ conda activate automatic-data-processing
$ conda install qgis --channel conda-forge
$ conda install -c conda-forge ocl-icd-system
$ pip install pvl pdr
$ pip install opencv-python-headless

File setup

First, clone the repo.

Please download the MSL and Mars2020 DEM files from this folder and place them into the qgis directory in your cloned repo: https://drive.google.com/drive/folders/1GmYnekSMn2mPa3q1FcFp7Cfd1eF9q1bA?usp=sharing

Place the Mastcam images that you'd like to have processed into qgis/msl_images and labels into qgis/msl_labels for Curiosity rover images, and qgis/mars2020_images and qgis/mars2020_labels for the Perseverance rover images. Make sure that the images you pull from NASA are MastCam or MastCamZ, and not another camera!

Running the code

In the qgis_pipeline.py script, first check that the QGIS_PYTHON_INSTALL path and the WORKSPACE_PATH are correct. Then, in the main function, make sure that MISSION is representative of your desired mission: either MSL() or MARS2020().

Then, run the code by executing the following in your conda environment: python3 qgis_pipeline.py.

To see the generated viewsheds, open up QGIS and open the base_msl.qgz or base_mars2020.qgz (depending on whether you are working with Curiosity or Perseverance images).

Troubleshooting

Feel free to post an issue if you have any difficulties, or email me at anjashep@umich.edu! There are comments in the qgis_pipeline.py file that may help you add debugging print statements to figure out any potential issues.