AFOSR’s investment in a data mining tool has transitioned from Basic Research to TRL6 in collaboration with Space Force Delta 2 just in two years! It’s relatively easy for people to identify clear outliers in clusters of data, but computers have difficulty with this. They have even more trouble with multiple clusters, such as satellites and debris in orbit. The problem of identifying outliers becomes more complicated when the data is already somewhat grouped, and when the items in the groups are slightly different. This project is aimed at identifying outliers under generalized conditions and developing an efficient data mining framework. 

Principal Investigator Dr. Sunil Aryal of Deakin University in Melbourne, Australia has developed an approach that can identify objects within clusters of space debris that are behaving differently from all the other objects. This could potentially identify nefarious actors attempting to obscure their activities among large clusters of objects. 

AFOSR’s investments in this technology comes from two grants: one to investigate robust data mining techniques, and a follow-on grant to adapt algorithms for use in detecting outliers and anomalous behavior in large heterogenous datasets. This investment equals $150 thousand between FY21 and 22. 

This technology has transitioned to the U.S. Space Force’s Space Operations Command Delta 2, who are currently testing the concept on space tracking datasets to identify anomalous behavior for individual objects among 50,000 items. 

“Transition of Practical Data Mining to the SpaceForce Operation Unit in such a short period of time is pretty much unprecedented,” said AFOSR Program Officer Dr. Geoffery Anderson. “Dr. Aryal’s project is developing new algorithms, which puts his work squarely in the basic research category, but for him to also be able to work on solving a mission-focused problem at the same time is rare. Typically, basic research investments take 20 years to get to that point.” 

The Basic Research Enterprise often takes a long time between initial investment and a useful tool, but this shows that investments can, at times, result in rapid transition.

Dr. Aryal is offering PhD Scholarships to strong domestic and international candidates in the areas of AI, ML, and data mining: https://sunilaryal.github.io/

Originally Published: September 7, 2023