Space has been a warfighting domain in practical terms for decades – GPS-guided munitions, satellite communications, overhead reconnaissance, and missile warning systems all depend on assets in orbit. What has changed in the past ten years is the explicit recognition within NATO and allied space commands that adversaries are actively threatening those assets, and that passive awareness of what is in orbit is no longer sufficient. Space domain awareness (SDA) software is the technical response: platforms that not only track the orbital environment but analyze it for hostile intent, attribute threatening behaviors to specific actors, and feed that picture into military command decisions.
This article covers the engineering architecture of SDA software – from the sensor networks and data fusion pipelines that build the orbital picture, to the orbit determination algorithms, conjunction analysis engines, and threat assessment workflows that turn raw observations into actionable intelligence. It is written for defense program managers, space operations engineers, and C2 architects evaluating or building SDA capabilities.
Why space is a contested warfighting domain
The low Earth orbit environment has become critically congested. Commercial mega-constellations – Starlink, OneWeb, and their successors – have added thousands of active satellites to a belt already crowded with decades of accumulated debris. The US Space Surveillance Network catalog now tracks approximately 27,000 objects larger than 10 cm; statistical models estimate over 500,000 objects larger than 1 cm that cannot be individually tracked but are large enough to destroy or disable a satellite on impact. This congestion creates collision risk as a constant background operational challenge even before adversarial activity is considered.
Against this backdrop, three categories of threat drive the military SDA requirement. Anti-satellite (ASAT) weapons – kinetic kill vehicles, directed energy systems, co-orbital interceptors – directly threaten high-value space assets. China's 2007 ASAT test and subsequent developments demonstrated that destroying satellites is a near-peer adversary capability, not a theoretical concern. Beyond kinetic threats, electronic warfare against space assets has proliferated: GPS jamming and spoofing is documented in multiple active conflict zones, and satellite communications jamming has been used to degrade secure communications links. Co-orbital threats – satellites maneuvering to close proximity with high-value assets for inspection, interference, or attack – are the most difficult to characterize because they exploit the same maneuver behaviors used for routine station-keeping and orbital maintenance.
The dependency chain amplifies the stakes. GPS underpins precision navigation for ground, air, and maritime forces. Satellite communications carry command traffic, ISR data, and coordination for distributed forces. Weather and reconnaissance satellites feed intelligence and planning workflows. Disrupting any of these services degrades joint force effectiveness in ways that compound across the operational area. SDA software exists to protect these dependencies by providing early warning and attribution before disruption becomes unrecoverable.
Key distinction: Space situational awareness (SSA) tells you what is in orbit and where it is. Space domain awareness (SDA) tells you what is happening in the space domain, who is responsible, and what it means for military operations. The shift from SSA to SDA reflects the recognition that passive tracking is no longer sufficient.
Sensor networks: the eyes of an SDA platform
No single sensor type can observe the full orbital environment. SDA platforms are inherently multi-sensor fusion systems, combining complementary observation modalities to achieve coverage across all orbital regimes.
Ground-based optical sensors range from commercial telescope networks to dedicated government phased-array electro-optical systems. Optical sensors observe objects in medium Earth orbit (MEO) and geosynchronous orbit (GEO) that are illuminated by sunlight against a dark sky – a geometry that requires observing from twilight periods when the ground site is in darkness but the target orbit is still sunlit. They provide high-precision angular measurements (right ascension and declination) but no range information directly, requiring multiple observations from different sites or across time to determine an orbit. Optical sensors cannot observe objects in LEO during most of an orbit because those objects pass through the Earth's shadow; they are also degraded by cloud cover and light pollution. The commercial space object observation market – companies like LeoLabs, ExoAnalytic, and AGI – has substantially expanded the optical observation network available to military SDA programs through data-sharing agreements.
Ground-based phased-array radars are the primary sensor for LEO objects. The US Space Fence on Kwajalein Atoll, operating at S-band, can detect objects as small as 2 cm in LEO and processes tens of thousands of observations per day. Earlier generation mechanical-dish tracking radars (FPS-85, GLOBUS II) are supplemented by newer electronically steered arrays that can observe multiple objects simultaneously without mechanical slewing delays. Radar provides range, range-rate (Doppler), and angular measurements – a richer observation type than optical angles-only, enabling shorter arc orbit determination with higher initial accuracy. Radar is weather-independent but horizon-limited: it observes objects within its field of regard, and global coverage requires a network of stations at geographically distributed sites.
RF collection systems monitor the electromagnetic emissions of space objects. Signals intelligence (SIGINT) receivers characterize the transmission signatures of active satellites – frequency, modulation, power, pulse characteristics – enabling identification and monitoring of changes that may indicate mode changes, anomalies, or new capabilities. RF interference monitoring detects jamming and spoofing events against GPS and satellite communications links, attributing interference to geographic source regions using direction-finding networks. When an RF interference event correlates with a maneuvering co-orbital object, the combined signature is a strong indicator of hostile action rather than a technical anomaly.
Space-based sensors – telescopes aboard satellites in GEO looking inward at the LEO belt – provide coverage in geographic regions where ground stations cannot be positioned and are not subject to atmospheric or weather degradation. The US Space-Based Space Surveillance (SBSS) program demonstrated this capability; allied programs and commercial equivalents are expanding the space-based sensor network. Space-based sensors also observe objects in GEO from a close-proximity vantage that enables finer characterization of object shape, attitude, and operational state than is achievable from ground stations at 36,000 km range.
Data fusion pipeline: from raw observations to orbital catalog
The SDA data fusion pipeline transforms heterogeneous sensor observations into a maintained catalog of orbital objects with associated state vectors, covariances, and threat classifications. Each stage of the pipeline has distinct engineering requirements.
Observation ingestion and normalization receives observations from each sensor in its native format and converts them to a common internal representation. Every observation carries sensor identifier, observation time (UTC to microsecond precision), measurement type and values, measurement noise covariance, and the sensor's own state vector at observation time. Precise time tagging is non-negotiable: a 1-millisecond timing error corresponds to roughly 7 meters of position error for a LEO object moving at 7.5 km/s. Sensor bias calibration – characterizing systematic offsets in each sensor's measurements – is performed periodically using observations of well-known calibration objects whose orbits are established with high precision.
Initial orbit determination (IOD) processes short observation arcs from new uncatalogued objects to produce a first estimate of the orbital state. Classical IOD algorithms – Gauss, Laplace, and Gooding methods – require a minimum of three observations to solve for the six orbital elements. The output of IOD is a preliminary orbit with large uncertainty; it is sufficient for the catalog to begin tracking the object but requires additional observations for operational accuracy. The IOD module also handles the association problem: determining whether a new observation arc belongs to a previously catalogued object or represents a genuinely new object. This is particularly challenging in the aftermath of fragmentation events that can create hundreds of new objects simultaneously.
Differential correction (orbit determination update) refines the orbital state vector by fitting accumulated observations using iterative least-squares or batch sequential estimation. The force model applied during propagation must accurately reflect all perturbations: atmospheric drag (critical in LEO below 800 km altitude, where even small density variations cause significant along-track drift), solar radiation pressure, Earth's non-spherical gravitational field (J2 through J6 harmonics), and third-body effects from the Moon and Sun. Real-time atmospheric density model updates – using geomagnetic index and solar flux data – are essential for maintaining LEO catalog accuracy during periods of elevated solar activity when atmospheric expansion significantly perturbs drag-dominated orbits.
Catalog maintenance and maneuver detection continuously monitors catalogued objects by comparing new observations against predictions propagated from the current element set. An object whose observed position deviates from prediction beyond the noise floor of the orbit determination process is flagged as maneuvering. The maneuver detection module initiates intensive observation re-tasking for the flagged object, suspends its conjunction screening (since its future orbit is now unknown), and triggers a maneuver characterization workflow to determine the delta-v applied and the resulting new orbit. Non-cooperative objects – military satellites from adversary nations – that maneuver without advance notice receive immediate threat assessment processing.
Engineering note: Atmospheric drag uncertainty is the dominant source of LEO catalog error during periods of elevated solar activity. A geomagnetic storm can increase atmospheric density at 400 km altitude by a factor of ten, advancing re-entry predictions by hours and degrading conjunction analysis accuracy across the entire LEO catalog until new observations are processed. SDA platforms must propagate drag uncertainty through to conjunction probability estimates, not treat drag as a deterministic perturbation.
Conjunction analysis: computing collision risk at scale
Conjunction analysis – identifying close approach events between tracked objects – is computationally demanding at catalog scale. Checking every possible pair of 27,000 objects against every future timestep at high fidelity is computationally infeasible in real time. Production SDA platforms use a hierarchical screening architecture that eliminates the vast majority of impossible conjunctions with cheap geometric tests before applying expensive numerical propagation to the small fraction of pairs that require it.
The first screening stage applies a geometric filter based on the minimum orbital separation between two objects' orbits (the distance of closest approach for the osculating orbit pair without considering phasing). Pairs whose minimum orbital separation exceeds the screening volume threshold – typically 5 km radial by 25 km along-track for LEO active satellites – are eliminated without further processing. This filter reduces the candidate pair count by several orders of magnitude. A second filter checks period compatibility: two objects in significantly different orbital periods will only be near each other infrequently, and if the next such time is beyond the screening window, the pair is deferred. Only pairs surviving both filters proceed to high-fidelity propagation.
High-fidelity propagation uses a numerical integrator (Runge-Kutta 4/5 or equivalent) with the full force model to propagate both objects forward to their predicted closest approach time. The state covariances are propagated simultaneously – using either linearized covariance propagation or Monte Carlo sampling – to compute the combined uncertainty ellipsoid at closest approach. Probability of collision is computed from the combined covariance and miss distance using analytical methods (Foster/Akella formula) or Monte Carlo integration for highly non-linear covariance shapes.
The output of the conjunction analysis pipeline is a Conjunction Data Message (CDM) for each event below the screening threshold. CDMs are distributed to satellite operators, space operations centers, and the common operational picture based on asset ownership and classification level. Military CDMs for high-value assets carry additional fields: threat classification (nominal conjunction vs. suspected proximity operations), recommended maneuver options with fuel cost estimates, and time-to-maneuver-decision deadline based on the time required to plan and execute an avoidance burn.
Threat assessment: from tracking to intelligence
The threat assessment layer is what distinguishes military SDA platforms from purely technical SSA systems. It applies intelligence tradecraft to orbital mechanics data to characterize adversary intent and provide commanders with assessments that support decision-making.
Maneuver attribution and pattern-of-life analysis builds behavioral baselines for catalogued objects. Every active satellite has a characteristic maneuver pattern: commercial communications satellites perform regular station-keeping burns to maintain orbital slot; reconnaissance satellites maneuver to adjust ground track phasing; debris avoidance maneuvers follow predictable geometries driven by CDM warnings. Deviation from the established pattern-of-life – an unusual maneuver magnitude, an unexpected orbit change, activity outside the normal maneuver season for the satellite type – triggers analyst review. Co-orbital maneuvering that places an object on a trajectory toward a high-value asset without operational justification is classified as a proximity operations event requiring escalating response options.
RF jamming attribution correlates RF interference detections with the orbital mechanics of known RF-capable objects. When a GPS jamming event is detected in a geographic region, the threat assessment module queries the catalog for objects with known RF payload capability whose ground coverage footprint includes the affected area at the relevant time. Correlation of orbital geometry with interference timing provides attribution confidence for the jamming source. Similar analysis applies to satellite communications jamming: the uplink jamming source is localized using direction-finding triangulation, and the threat assessment module correlates the source location with known ground-based electronic warfare assets in the relevant adversary's order of battle.
Re-entry prediction and debris risk assessment becomes a threat assessment function when the re-entering object is a weapons delivery vehicle or when the re-entry trajectory could be misinterpreted as a ballistic missile launch by missile warning systems. SDA platforms maintain re-entry predictions for all decaying LEO objects, with uncertainty bands that narrow as re-entry approaches. For objects with re-entry windows over populated areas or strategically sensitive regions, the threat assessment layer generates advance notifications to civil defense authorities and missile warning operators to prevent misclassification.
Space object COP layer: integrating SDA with military C2
The operational value of SDA software is realized when the space picture is integrated into the same command and control environment used by commanders for all other force elements. A standalone SDA display that space operators monitor in isolation cannot inform joint force commanders in time to affect decisions about GPS-dependent operations, satellite communications routing, or ISR collection planning.
The space object COP layer publishes the space catalog as a distinct track layer within the joint COP, accessible alongside terrestrial air, land, and maritime tracks. Space object tracks carry orbital parameters, conjunction warning status, threat classification, and attribution data as track attributes. The visualization presents a 3D orbital display showing the main orbital shells (LEO, MEO, GEO), active conjunction events as geometric overlays between converging orbital paths, and threat classification color-coding that immediately communicates which objects are under active assessment.
Mission planning integration surfaces space asset availability to operational planners. GPS availability analysis – showing ground coverage quality as a function of satellite geometry and known jamming environments – is computed and displayed as a time-varying overlay on the operational map. Communications satellite window planning identifies blackout periods when relay satellites pass below the horizon for specific ground terminals. ISR satellite overflight timing is integrated into collection management planning. These functions require the SDA platform to feed its space object catalog and threat assessments into the same data fabric that underpins the common operational picture, not to operate as an isolated specialty system.
For multi-domain C2 environments built on platforms like Corvus.Head, the space COP layer integrates through the same track ingestion and correlation pipeline used for other sensor-sourced tracks. Space object tracks use standard track message formats with space-domain-specific extensions for orbital elements and conjunction data. This allows space operations staff to work within the same interface as the rest of the joint operations center while space-specific analysis tools are accessible within the same platform rather than requiring context-switching to a separate application.
Integration requirement: The space COP layer must present space asset status in terms operationally meaningful to non-space commanders – GPS availability quality, communications satellite coverage windows, ISR overflight timing – not raw orbital mechanics data that requires specialist interpretation. The translation from orbital mechanics to operational impact is a software function, not a task for already-overloaded operations staff.
Software architecture: orbit propagation, kalman filtering, and 3D visualization
The computational core of an SDA platform combines two distinct performance regimes: batch processing for catalog maintenance and conjunction screening, which can tolerate latencies of minutes to hours; and real-time display and alert delivery, which must operate at sub-second response times for the operator interface and near-real-time for warning dissemination.
SGP4/SDP4 – the Simplified General Perturbations models published by the US Space Track program – remains the standard for rapid catalog propagation and for publishing element sets that can be consumed by downstream users without requiring access to the proprietary force models of the originating sensor network. SGP4 is analytically tractable (propagation of a single object requires microseconds of CPU time) and produces position predictions accurate to 1–3 km over a 24-hour propagation window for typical LEO objects. For conjunction analysis and precision maneuver detection, higher-fidelity numerical propagators are used that incorporate real-time atmospheric density, detailed solar radiation pressure models, and higher-order gravitational terms – at the cost of significantly higher computational load.
Sequential orbit determination uses extended Kalman filtering (EKF) or unscented Kalman filtering (UKF) to process observations as they arrive and update the state estimate incrementally, rather than waiting for a full observation arc to run a batch least-squares fit. The UKF is preferred for highly non-linear observation geometries – angles-only observations from a single site, short-arc initial orbit determination – where the linearization approximation of the EKF introduces significant errors. The covariance matrix maintained by the filter is not merely a byproduct of the estimation process; it is a first-class data product that feeds directly into conjunction probability computation and determines the observation tasking priority (objects with high position uncertainty get scheduled for new observations sooner).
The 3D orbital visualization requires a specialized rendering architecture distinct from the 2D map rendering used for terrestrial COP displays. Orbital mechanics requires accurate representation of elliptical orbits at scales ranging from a few kilometers (conjunction geometry) to tens of thousands of kilometers (the full GEO belt). WebGL-based orbital viewers can render tens of thousands of object tracks at interactive frame rates using GPU-accelerated orbit propagation – computing ground tracks and orbital positions in the vertex shader rather than on the CPU. Time-acceleration controls allow operators to fast-forward through predicted conjunction events and re-entry windows, visualizing the orbital geometry at the critical time rather than waiting for it to arrive in real time.
How to build the data fusion pipeline for a space domain awareness platform
The following structured process translates the architectural principles above into a concrete design workflow for an SDA data fusion pipeline, from sensor ingestion through threat assessment and COP integration.
- Define the sensor network and observation data model – specify sensor types, native data formats, and design a normalized observation schema with sensor identifier, UTC timestamp (microsecond precision), measurement type, values, noise covariance, and sensor state vector. Time tagging precision and sensor bias characterization at this stage prevent systematic errors from propagating through orbit determination.
- Implement observation ingestion and quality control – build format adapters for each sensor's native output (OBSM, TDMF, proprietary radar formats), apply QC filters to detect anomalous measurements, and quarantine failed observations for analyst review rather than silent discard. QC failures during contested operations may indicate jamming rather than sensor malfunction.
- Build the orbit determination engine – implement initial orbit determination (Gauss or Gooding method) for new objects, and differential correction using iterative least-squares or UKF for catalogued objects. Select force model fidelity appropriate to the orbital regime: full drag, SRP, and gravitational harmonics for LEO; SRP and tesseral harmonics for GEO. Apply real-time atmospheric density updates for LEO objects.
- Implement the space object catalog – design the catalog data model (state vector with covariance, TLE, observation history, classification, attribution, operational status), build the update pipeline, and implement object identity resolution to distinguish new objects from re-acquired known ones. Log all maneuver detections, breakup events, and new object discoveries for intelligence reporting.
- Build the conjunction analysis pipeline – implement hierarchical screening (geometric filter, period filter, high-fidelity propagation), compute Conjunction Data Messages with probability of collision and miss distance, and build the automated CDM distribution layer routing warnings to satellite operators and the joint COP based on asset ownership and classification.
- Implement maneuver detection and threat assessment – build the residual monitoring module that flags deviating objects as maneuvering, integrate pattern-of-life analysis for anomaly detection, and implement the proximity operations classification workflow that distinguishes routine station-keeping from co-orbital threat behavior. Correlate maneuver events with RF collection data for attribution enhancement.
- Integrate with the military COP – publish space object tracks to the joint COP using standard track formats with space-domain extensions, implement GPS availability and communications window overlays for operational planners, and deliver conjunction warnings and threat assessments through the same alert architecture used for other intelligence products. Ensure space asset status is expressed in operationally meaningful terms, not raw orbital parameters.