Sandia's 12-foot SOHBRIT observatory tracks up to 80 satellites per clear night using commercial off-the-shelf hardware and an automated scheduler, generating image sets that feed detection-algorithm and machine-learning work at the national laboratory.

The facility sits in Albuquerque, N.M., and supports experiments and testing for Sandia's Space Mission Program. Operator Allen Schnibben, a computer scientist, sets nightly schedules from his desk and reviews a fresh dataset the next morning.

"From my desk, I can set up a schedule that will execute during the night," Schnibben said. "All I have to do is just come in the next day and I have a fresh new set of data that was collected overnight."

What's driving the workload?

NASA counted more than 45,000 human-made objects in Earth orbit as of 2024, a tally that has climbed sharply alongside commercial mega-constellations. The trend determines both the volume and the variety of objects Sandia's telescope must capture.

"Satellites travel very fast," Schnibben said. "So, a collision at the speeds you expect satellites to be at could be a very catastrophic collision." Debris from any such event would multiply hazards for the rest of the constellation, raising the stakes for routine tracking.

University teams now build small payloads and reach orbit, a shift that broadens the catalog of objects requiring characterization. "Space is becoming more readily accessible for different groups of people," Schnibben noted, listing communications, imagery and academic missions as drivers of orbital density.

How was the system built?

Sandia constructed SOHBRIT entirely from commercial off-the-shelf products. The intent, Schnibben said, was to assemble a system that is "relatively low cost" and surround it with automation software so that "you don't need a person out there to operate the observatory."

The design separates the project from custom astronomical instruments typical of large government observatories. R&D managers weighing similar ground assets can compare SOHBRIT's throughput — roughly 80 satellites per clear night, by Schnibben's account — against the acquisition and staffing costs of purpose-built optical systems.

What does the output support?

The collected imagery feeds downstream experiments in detection algorithms and machine-learning models at the lab. "It takes images of these satellites and then Sandia employees can use that collective imagery to then do experiments on things like detection algorithms or maybe machine learning models," Schnibben said.

Such datasets matter for pushing limits on what optical systems can resolve, particularly as objects dim with altitude or shrink with miniaturization. "It helps with things like predicting if satellites are going to collide with each other or if one satellite has moved," Schnibben added. "Essentially, it's keeping track of where all these satellites are, which is an important mission for the military and Sandia."

What should R&D leaders interrogate?

Schnibben frames the work as image collection for "experiments." The source does not cite published, peer-reviewed results, leaving open whether detection-algorithm outputs have appeared outside Sandia's internal channels. Program managers evaluating similar initiatives typically want that trail documented before counting the work toward a portfolio.

The 12-foot aperture and off-the-shelf hardware also imply a sensitivity ceiling relative to dedicated deep-space tracking assets, a constraint worth flagging when extrapolating machine-learning performance across dimmer or smaller object classes.

Clear-night uptime determines dataset continuity. The source does not address seasonal gaps from New Mexico weather patterns or backup observing capacity, both standard inputs for an R&D portfolio review.

What should managers track next?

The project points to a procurement template: low-cost hardware, automated scheduling, and research output oriented toward algorithmic rather than optical innovation. Three measurable signals worth monitoring:

• Published detection-algorithm benchmarks that cite SOHBRIT imagery
• Throughput beyond the 80-satellite-per-night target under varied sky conditions
• Integration with space-object catalogs outside the current mission set

Schnibben indicated further work will examine whether machine-learning techniques can extract more signal from the same COTS hardware — and whether mission requirements will eventually justify upgrades to higher-sensitivity instrumentation.