Undergrad students deploy applications to geosynchronous satellite 22,236 miles above Earth

For many college students, satellites are weak points crossing the sky on clear nights. These objects are background details from science fiction to reality, such as GPS, satellite radio and Wi -fi.

For students of computer engineers of the University of Pittsburgh, Dikhya and Nishal Harela, compact objects that live through space and abandoned by radiation are unique problems.

Thanks to partnerships with the center of space, high -performance and sustainable computing of the University of Pittsburgh and Lockheed Martin, two students had a unique opportunity to develop and deploy applications for a satellite in Geosinchronous orbit (GEO), 22 236 miles above the ground.

Detection of a spike

Since the cousins ​​approached the end of the engineering school Svanson last spring, not one of them had specific plans to visit graduate school. This changed after a conversation with Samuel Diexon, an assistant professor at the Department of Electric and Computer Technology.

“When Professor Dilerson found out that we both loved space and want to work in NASA, he asked if we knew about Shrek,” the niche said. “In some way we did not do this.”

Dilexon changed this, presenting them with Alan George, the Department of the Department, R & H Mickle andowed Chair, director of the NSF space center, and professor of electric and computer engineering. George provided Dikhya and Nishala with the opportunity to earn loans, conducting space research with Linus Silbernagel, second -year doctors of philosophy. Student, and Evan Gretok, post -instrument researcher Schrek.

Solving real problems

On November 1, 2022, Lockheed Martin launched a satellite system of updating in space (LM Linuss) in GEO, which allows the satellites to the orbit at the same speed as the rotation of the earth, which makes them stationary from the ground. Today, when the mission is approaching the end, the Cubesat satellite serves as a test bench.

In GEO, more than 20,000 miles further than satellites in low -land orbit (LEO), Linuss provides the opportunity to test the code and applications in more difficult conditions.

“Lockheed Martin turned to Dr. George about the possibility of creating and conducting tests on the GEO satellite from the SHREC experience in the field of satellite equipment and software,” said Silburnagel. “We had some problems that dicha and a niche could help solve.”

Memory limits

Cubesat satellites are incredibly small – less than four inches – and thus limited in their computing power, reducing the effectiveness of the function. For example, tiny cameras can capture the lengths of waves that the human eye cannot process, but creates huge files that spend time and energy to transfer back to Earth.

“I worked on a project called CNN JPEG, which began a graduate student, and which uses machine training to compress data in satellite photographs,” the niche said. “I wanted to deploy the technology on Linuss.”

The Linuss satellite uses ZCU102 computing board, and the team has reproduced Lockheed Martin satellite configuration on it.

“Lockheed Martin packs its applications in a certain way,” the niche said. “We received a technology working on our test board, and sent it to them to transfer to Linuss.”

Unfortunately, although the application worked on Earth, it was too big to run into satelliteThe field “In the classroom we are taught about the restriction of memory, but this differs when you actually encounter this,” said the niche, which works on optimization of CNN JPEG.

Forcing him to work in space

Satellites generate many images, but most do not matter. They depict areas of the ocean or forest or just clouds. “The satellites send many images that are of little interest for researchers, spending memory and bandwidth, which creates delays,” said Dikhhya.

To solve this problem, Dikhya decided to run Machine training model On Linuss. The model trained in Greek provides an autonomous built -in classification on a limited system. The application reduces mass images in tiles and classifies them to help eliminate excess data. “My work was to force him to work in space,” Dikhya said.

She did just that.

Using the ZCU102 fee, she programmed and packed the application and sent it to Lockheed Martin.

“Mine is a smaller model, so she ran without problems,” said Dikhya. “After our last tests, the Lockheed Martin team sent it to Linuss, launched the application, and the model correctly classified the images. View what we built successfully worked 22,000 miles above the Earth was an important milestone and exciting opportunity. ”

“It would be impossible for students to do something so big without Lockheed Martin without partnership,” said Silburnagel. “Dikhia and the niche remotely joined the team during testing and deployment. For the first time, Pitt students manage applications in Geo. ”

“This project is a significant aspect of Shrek,” George said. “Through partnerships at the university and industries, like this, students receive these incredible opportunities that promote them in Space research And prepare them for a career in space engineering. ”

“When we started,” said the niche, which, like Dikhya, is now a SHREC graduate student, “we were students with practically without real research experience. We were faced with problems that others did not have, and we had to find out how to solve them. ”