Quantum-inspired software shrinks orbital tracking models by 99%

A New York technology company has won its first U.S. federal research contract to develop software that could help military operators identify unknown objects in Earth’s orbit more quickly. The award comes through the SpaceWERX Open Topic Small Business Innovation Research (SBIR) program and focuses on improving Space Domain Awareness using physics-constrained machine learning.

BosonQ Psi Federal (BQP) will validate a new software application that combines physics-based models with quantum-inspired computing techniques. The goal is to classify unidentified orbital objects faster while using significantly less computing power than conventional artificial intelligence models. The technology targets satellites and other edge platforms that operate with strict power and processing limits.

Crowded orbital environment

The growing number of satellites and debris has made tracking activity in orbit increasingly difficult. The U.S. Space Surveillance Network collects between 18,000 and 25,000 observations every day. Many of those detections cannot immediately be linked to known satellites or debris.

These unidentified observations, known as Uncorrelated Tracks (UCTs), can include newly launched spacecraft, collision fragments, or objects that require closer investigation. Delays in identifying them can slow operational decisions and reduce overall awareness of the space environment.

BQP’s software aims to speed up that process by combining physics constraints with quantum-assisted machine learning. The company says the approach delivers accurate AI inference without depending on cloud infrastructure, graphics processors, or future quantum computers. Instead, it is designed to run directly on space-qualified processors and other resource-constrained hardware.

Smaller AI, faster decisions

According to BQP, its Physics-Constrained Quantum-Assisted Machine Learning (PC-QAML) architecture produces models that are 99% more compact, shrinking them from roughly 14 million parameters to about 2,000 without sacrificing accuracy. The company says the software maintains more than 99% classification accuracy despite the dramatic reduction.

The compact architecture also delivers up to a tenfold reduction in inference latency and lowers power consumption by roughly 90%. BQP says engineers can retrain the models significantly faster than conventional machine learning systems.

Those efficiency gains have already enabled deployment on an NVIDIA Jetson Nano edge computing device at the Space Domain Awareness TAP Lab, formerly known as the SDA TAP Lab. The demonstration showed that advanced AI could operate on compact hardware suitable for autonomous space missions.

“Our goal is to make advanced AI practical where it matters most: on satellites and forward-deployed systems operating with limited computing power and intermittent communications,” said Rut Lineswala, founder and chief technology officer of BQP.

He said the federal award validates the company’s technology and provides an opportunity to demonstrate how quantum-inspired computing can address operational challenges facing national security missions.

Beyond military missions

Military operators could use the software to distinguish routine orbital activity from potentially suspicious behavior, including satellite maneuvers, separation events, and close-proximity operations. Processing information directly on spacecraft could reduce dependence on centralized computing systems and improve response times.

The project expands work BQP previously carried out with the Space Domain Awareness TAP Lab. During the 2025 SDA Mini-Accelerator, the company’s technology demonstrated orbital separation detection capabilities and emerged as a candidate for future UCT classification and threat simulation applications supporting U.S. space operations.

Recommended Articles