How do humans behave off-planet? In order to answer this question, one must have a variety of data sources which are then analyzed to reveal behavioral patterns. Amir Kanan Kashefi works at the intersection of archaeology, space, and data science. He compiles and makes sense of data regarding astronauts’ behavior on the International Space Station. We spoke with him to learn more about the data sources, their analysis, and the purpose of this project.
What space-related data sources do you deal with?
I’m the data engineer for the International Space Station Archaeological Project (ISSAP), which is the first large-scale analysis of an archaeological site in space. The International Space Station has been continuously occupied for 20 years now, and an archaeological approach has never been taken before. ISSAP is a collaboration between Flinders University in Adelaide and Chapman University in California, with data provided by NASA. I work closely with the principal investigators on the project, Associate Professors Justin Walsh and Alice Gorman. What we’re aiming to do is use the NASA image archive to study how the crew interact with objects and spaces inside the ISS and how this interaction changes over time. We can use our results to help design future missions and habitats.
We have four sets of unstructured and structured data. The unstructured data come from archived photographs taken by astronauts while they were working and living in space. In addition, ISSAP utilizes the public multimedia dataset of ISS images, which have been uploaded to NASA Johnson’s collections on Flickr. These images are enriched with some metadata such as date, location, and astronauts’ names that need to be extracted. The structured data are the ISS Inventory Management System (IMS), a legacy relational database management system. Space agencies use the IMS to keep track of items sent to ISS. Lastly, ISSAP observes the items returned from ISS to Earth by analyzing the cargo documents.
How do you help analyze these data sources?
The NASA image archive is analyzed by subfields of artificial intelligence such as machine learning, neural networks, deep learning, image segmentation, and computer vision. ISSAP applies these techniques to analyze how crew members interact with objects and space on the ISS. To address questions of this kind, entities such as people, objects, locations, and events need to be identified. These entities are highly linked and the relationships (links) between them must be stored in a database along with actual data. Graph analytics have been used to analyze the patterns hidden in our unstructured data.
To clean a legacy database like IMS, we inspect the database to figure out what data exist and understand the structure of it. The inspected data have been categorized to three data groups: (1) the data that can be used unmodified, (2) the data that must be converted, and (3) the data that cannot be used at all. We use data classification techniques to understand the item types, proportions, and purpose of each item. This process starts by writing different queries to search and assign a sub-item to its ancestor (e.g., laptop to electronics), using our data model of nearly 70 entities and sub-entities.
What are the final end products that you are enabling?
As a product, we are designing and developing a hybrid database that is an innovative integration of the relational and graph databases. This database enables us to store and augment the NASA image analysis results along with the corresponding data from IMS tables and cargo spreadsheets. Moreover, we have developed an ontology that is the backbone of our database and applies space archaeology knowledge to the data described. This forms a knowledge base that contributes insights to the design of future long-duration space missions and to maximizing both survival and efficiency. We envisage that a scientist or mission designer will be able to search for a particular item in the database and see how astronauts have used it (or have not used it) over time. This might be something as simple as a zip-lock bag, of which there are thousands, or a pair of scissors.
This project fills in the gaps in other scientific research (e.g., in social science and engineering) by providing information about norms, symbols, and artifacts in the unique space environment.