STELLAR SCIENCE LTD. CO. — Department of Defense SBIR Phase I: HQ085021S0001-07

Description

The Space Development Agency (SDA) is developing the National Defense Space Architecture (NDSA), which entails a national defense mission to provide constant global surveillance for military threats. However, picking targets out of a noisy, changing, and varied background and tracking them requires advanced algorithms and system designs combined with extensive testing. We propose to develop a detection and tracking algorithm based on an infrared-path image (IPI) model that has achieved state-of-the-art performance in target identification and augment it with a novel multi-satellite tracking component appropriate for a proliferated low Earth orbit (LEO) architecture. The IPI model is based on newly developed robust principal component analysis (RPCA) techniques, where a convex optimization separates a low-rank background from sparse foregrounds. Objects detected in this sparse foreground across multiple platforms in the LEO architecture will feed multi-target tracking data correlation and Kalman filtering algorithms. By developing these in a way appropriate for an open system architecture, they will be readily available for transition to a real-world system. Architectural performance is not based on single-sensor detection receiver operating characteristic (ROC) performance, but rather how well the proliferated architecture can perform. Ensuring any algorithm developed by us or others meets design goals for the constellation requires extensive testing and validation. No existing framework is currently adequate for this task. Alongside the detection and tracking algorithm, we propose to develop Argus, a robust and accurate modeling and simulation framework designed to evaluate these complex systems. Argus will be built on top of the Advanced Framework For Simulation, Integration, and Modeling (AFSIM),  the trusted DoD simulation framework, allowing for the entire NDSA constellation to be tested in conjunction with a multi-domain engagement. Using a plug-in infrastructure, combined with the open system architecture used for our algorithm, will allow many algorithms to be tested, even if developed by others. For imagery, Argus will leverage a radiometrically accurate ray tracing scene generation tool that includes realistically modeled terrain (sea, land, and coasts) backgrounds. This unmatched combination of flexibility and accuracy will enable testing on a scale and with reliability otherwise impossible. Stellar Science is a software development company specializing in advanced algorithm development, with deep experience in space domain modeling and simulation, radiometrically accurate image simulation, image processing, computer vision, and more. Argus will combine this expertise to meet and exceed SDA’s requirements for testing proliferated satellite architectures as described for Phase I. Its flexibility and robustness will enable it to fulfill program needs for Phase II and beyond.