Space Domain Awareness & Space Superiority

We’ve engineered cross-domain SDA solutions for critical National programs. Our mission expertise spans Space-Based and Ground-Based Situational Awareness.

SDA RPO Activity Prediction for Anomalous Behavior Classification

RPOs can be either benign mission activities or potential threats. Understanding the difference based on pattern-of-life anomaly detection and comparisons to prior/known behaviors drive earlier delivery of indication and warning (I&W) to the operator to enable faster and more decisive reaction decision making.  New ARKA technology automatically classifies new, anomalous behavior as threatening, cooperative, or inadvertent.

Lumos Sensor Data Quality Analyzer

Automated processing of incoming sensor data measurements is sometimes conducted within a sensor suite, but almost never across disparate systems providing the data to other downstream SDA processes.  Lumos solves this mission shortfall by sensing sensor biases and alerting operators to potential anomalies before that data is disseminated.  Additionally, if sensors are determined to be operating within known parameters, the same anomalous data alerts can be assessed to detect unplanned object maneuvering. Autonomous measurement processing alleviates operator burden and reduces latency for producing updated conjunction assessments, while providing realistic uncertainties for accurate collision probabilities.

Lumos Collection Recommender for Resource/Sensor Task Optimization

Optimizing the collection opportunities for highly valued assets is a complex process that involves the best lighting, look angles and sensor-to-object geometry.  Lumos automates this optimization process for operators and enables the pre-planning of collections on objects of interest and the prediction of future co-variance to acceptable mission levels.  Real-time optimization and deconfliction of tasks allows for ad-hoc/high-priority interruptable tasking and minimized mean-time-to-revisit (MTTR) for confident space object custody.

SDA Rendezvous & Proximity Operations (RPO) Automates Activity Classification and Prediction

Models that sense and predict RPO behaviors based on pattern-of-life anomaly detection and comparisons to prior/known behaviors drive earlier delivery of indication and warning (I&W) to the operator to enable faster and more decisive reaction decision making.  Our technology automatically classifies new, anomalous behavior as threatening, cooperative, or inadvertent.

Electro-optical (EO) imagery right-ascension and declination angle detections are used to characterize standard and abnormal behaviors.  Anomalous behavior can trigger indication and warning (I&W) alerts to acquire additional data (i.e., watch more closely) and initialize decision support options for possible reaction.

The machine learning (ML) Transformer Encoder-based AI model is initially trained with observations from several real-world data scenarios.  Then, real-time actual data is used to discern a normal/standard pattern-of-life.  Deviations from normality cause the model to automatically reclassify the object(s).

Modern software factory approaches to SDA maximize efficiencies for high-performance. ARKA’s RUST-based propagator is an example of this, as it is designed to efficiently propagate large space object catalogs by employing modern approaches to memory maintenance and multi-core/multi-thread operation.  The approach particularly shows benefit when propagating a large-number of space objects, e.g., full catalog propagation, by illustrating a >10x runtime improvement over a multi-threaded C# implementation.

Web assembly modules (WASMs) allow client-side compute for object propagation for low-latency, high-performance, seamless user interface experiences.  With this, the computationally intensive propagation burden is no longer server-bound and eliminates the need to relay large ephemeris data files to the client.  Instead, only state vectors (and covariance data) need be passed, allowing the client to decisively only propagate objects-of-interest as needed, which reduces or eliminates extraneous propagation computation.

Lumos is an integrated suite of capabilities and services created for USSF to improve and automate SDA sensor collections and data quality.  The Lumos Collection Recommender provides real-time, automated multi-sensor optimization for coordinated collection across multiple ground and space-based sensors.  The geometric diversity of the pre-planned observations greatly reduces uncertainty of RSO states and allows operators to balance mission requirements for timeliness, accuracy and availably of resources.  Based on mission objectives for a desired RSO covariance, a selectable algorithm such as Markov Chain Monte Carlo (MCMC) or Genetic Algorithm (GA) exposes the optimal coordinate collections for the RSO of interest.

Timely, low-latent collection allows for robust track custody and more confident positive object identification (POI) and related probabilities of association.
Through the use of a multi-state Unscented Schmidt Kalman Filter, the Lumos SDA Data Quality Anomaly Detector monitors real-time incoming radar and optical sensor data to determine if sensor biases are present and out of tolerance with known biases.  Operators are alerted to sensor biases, providing the means to prevent that data from entering downstream orbit determination tools and algorithms. These alerts can be indications of a change in object behavior (e.g., a spacecraft maneuver), or a change in the observing sensor’s mode.

Such messages can also be used to automatically trigger a sensor white-noise variance and bias parameter update, or to request further sensor re-calibration diagnostics.  Rapid detection, automated response, and alerting all work together to promote bringing the best data forward more rapidly to aid timely decision making.