In a significant demonstration of undergraduate research capabilities, two students from the University of Pittsburgh have successfully developed and deployed sophisticated software applications to a satellite orbiting 22,236 miles above the planet’s surface. The achievement marks a critical step in testing new methods for processing data directly in space, a capability that could redefine the efficiency of satellite communications and operations. The students’ work addresses a growing challenge in space-based systems: the limited bandwidth and computational power available for transmitting vast amounts of data back to Earth.
Working in collaboration with the university’s NSF Center for Space, High-Performance & Resilient Computing (SHREC) and aerospace giant Lockheed Martin, cousins Dikchhya and Nischal Kharel developed machine-learning applications for the In-space Upgrade Satellite System (LM LINUSS). This orbital platform serves as a testbed for new technologies in the harsh environment of geosynchronous orbit. Their software, designed to autonomously classify and compress satellite imagery, represents a hands-on application of advanced computing principles to solve real-world problems in astronautics, showcasing the impact of partnerships between academic institutions and industry leaders in fostering the next generation of space engineers.
A Serendipitous Path to Space Research
The journey for the two students from the Swanson School of Engineering into the realm of space technology was not pre-planned. As Dikchhya and Nischal Kharel neared graduation, they were uncertain about their postgraduate paths. A pivotal conversation with Samuel Dickerson, an associate professor of electrical and computer engineering, connected them with Alan George, the director of the SHREC Space Center. This introduction led to an invitation to participate in high-level space research projects, an opportunity typically reserved for graduate students.
Under the guidance of George and mentored by Ph.D. student Linus Silbernagel and postdoctoral researcher Evan Gretok, the Kharels were integrated into a team focused on practical space applications. This mentorship structure provided them with the framework and expertise needed to translate complex theoretical knowledge into functional software ready for the rigors of space. Their involvement began in earnest following the launch of the LM LINUSS system, providing them a rare chance to work directly with an active orbital asset and contribute meaningfully to an ongoing space mission.
The Orbital Testbed
The platform for their software experiment was Lockheed Martin’s In-space Upgrade Satellite System, which was launched on November 1, 2022. The system consists of two CubeSats, which are small, modular satellites, deployed into a geosynchronous orbit (GEO). This specific orbit is notable because the satellite’s orbital period matches the Earth’s rotational period, causing it to appear stationary from the ground. This characteristic is ideal for communications and surveillance but presents a challenging environment for electronics due to higher radiation levels compared to lower orbits.
Unlike satellites in low Earth orbit (LEO) that circle the planet every 90 minutes, GEO satellites maintain a fixed position, which is advantageous for continuous observation and communication. However, the immense distance imposes significant communication delays and data-rate limitations. The LM LINUSS satellite was specifically designed to serve as a versatile testing platform, allowing for the remote installation and execution of new applications. This “software-defined” satellite model enables engineers to trial new algorithms and processing techniques on orbit, accelerating the pace of innovation without the need to launch new hardware for every experiment.
Tackling Data Overload from Orbit
A primary bottleneck for modern satellites is the transmission of data. Earth-observing satellites, for example, can capture terabytes of imagery, but the bandwidth available to send that information to ground stations is extremely limited. A significant portion of the collected data, such as images of cloud cover or empty oceans, may be of little value. The Kharels’ work focused directly on mitigating this issue by making data processing more intelligent and efficient before transmission.
Onboard Image Classification
Dikchhya Kharel developed a machine-learning model designed to autonomously classify images captured by the satellite. Her application analyzes images as they are collected and determines their relevance, thereby reducing the amount of redundant or useless data sent back to Earth. By sorting and prioritizing images onboard, the system ensures that the limited communication bandwidth is used for the most valuable information. After extensive development and testing on the ground, her application was successfully uploaded and executed on the LM LINUSS, a major accomplishment for an undergraduate researcher.
Machine Learning for Data Compression
Nischal Kharel’s project targeted the same data bottleneck but through a different approach: data compression. He worked on a machine-learning application to compress satellite image data more effectively than traditional methods. One of the significant hurdles he faced was the inherent computational constraints of CubeSat hardware. These small satellites have limited memory and processing power compared to terrestrial computer systems. While his initial application showed success in ground-based simulations, it encountered memory allocation issues when first tested for the satellite’s systems, a common challenge when transitioning software from a lab environment to resource-constrained flight hardware.
From Terrestrial Success to Orbital Deployment
The process of preparing software for space is meticulous. Every line of code must be robust, efficient, and capable of operating with minimal resources. The students rigorously tested their applications to ensure they would function correctly on the satellite’s specific computing architecture. The successful deployment of Dikchhya Kharel’s image-classification software marked a key milestone for the entire project team. It validated the viability of uploading and running complex, student-developed code on a sophisticated commercial satellite in a high-orbit environment.
This achievement represents a significant educational and technical success. It provided the students with unparalleled hands-on experience in the full lifecycle of space software development, from initial concept and coding to validation and orbital execution. The project pushed the boundaries of what is considered possible for undergraduate researchers and demonstrated a functional model for how to test and de-risk new technologies in orbit, thereby lowering the barrier to innovation.
Bridging Academia and Industry
The success of this initiative underscores the immense value of strategic collaborations between university research centers and industry partners. The partnership between the University of Pittsburgh’s SHREC and Lockheed Martin created a unique educational pipeline, allowing students to work on cutting-edge technology and contribute to active space missions. Such programs provide students with invaluable, real-world experience that is impossible to replicate in a classroom setting, preparing them for future careers in the aerospace and technology sectors.
For the Kharel cousins, the experience was transformative. What began as an unexpected opportunity has now solidified their career paths; both are now pursuing graduate studies at SHREC, where they continue to contribute to space-related research. Their accomplishment not only represents a personal and academic milestone but also serves as a powerful example of how such collaborative initiatives can accelerate technological progress and cultivate the specialized skills required to advance the future of space engineering.
