Why NASA Is Now Using Floating Space Junk to Navigate Without Any GPS

                              Z (North Pole)
                                    ^
                                    |     Orbit of Tracked Debris (RSO)
                                    |     /
                                    |    /
                                    |   /
                                    |  o  r_tgt(t)
                                    | /
                                    |/
                                    +-----------------> Y
                                   / \
                                  /   \
                                 /     \
                                /       o  r_obs(t) (Observer Satellite)
                               /         \
                              /           \  Orbit of Observer
                             v             \
                            X

When an observer satellite at position $\mathbf{r}_{\text{obs}}(t)$ observes a piece of debris at position $\mathbf{r}_{\text{tgt}}(t)$, the measured unit line-of-sight vector $\mathbf{u}(t)$ in the camera frame is:

$$\mathbf{u}(t) = \frac{\mathbf{r}_{\text{tgt}}(t) - \mathbf{r}_{\text{obs}}(t)}{\|\mathbf{r}_{\text{tgt}}(t) - \mathbf{r}_{\text{obs}}(t)\|}$$

Because the debris is at a finite distance (ranging from tens of kilometers to a few thousand kilometers), the relative position vector exhibits substantial dynamic parallax. Over time, as both the host satellite and the space debris move along their respective orbital arcs, the measured angle $\mathbf{u}(t)$ sweeps across the camera detector at a non-linear rate.

If the orbit of the tracked debris object is cataloged—meaning its six orbital parameters ($a, e, i, \Omega, \omega, M$) are approximately known—the measured bearing angles over time can only be produced if the observer is situated at one unique trajectory through spacetime.

The Angles-Only Inverse Problem

Extracting both the host spacecraft's trajectory and the debris target's precise orbit from bearing angles alone is a non-linear inverse estimation problem. In astrodynamics, classical angles-only Initial Orbit Determination (IOD) algorithms—such as the Gauss method, the Laplace method, or the Gooding iteration—were formulated to estimate a single target’s orbit from a known ground-based observing site.

In the FALCON paradigm, the problem is inverted and compounded: neither the observer’s position nor the target’s position is assumed to be known with absolute precision.

+--------------------------------------------------------------------------+
|                  ANGLES-ONLY INVERSE PROBLEM FORMULATION                 |
+--------------------------------------------------------------------------+
|                                                                          |
|  Forward Physics Model:                                                  |
|  [State Vectors: X_obs(t), X_tgt_i(t)] ====> [Predicted Pixel LOS u_i(t)]|
|                                                                          |
|  Inverse Flight Estimation:                                              |
|  [Observed LOS Angles u_i(t)] + [Catalog Prior]                          |
|         |                                                                |
|         v                                                                |
|  Minimize Residual Cost Function:                                        |
|  J = SUM || u_measured(t_k) - u_predicted(X_obs, X_tgt, t_k) ||^2        |
|         |                                                                |
|         v                                                                |
|  Yields refined 12-state solution (6-state Observer + 6-state Target)    |
|                                                                          |
+--------------------------------------------------------------------------+

The onboard software executes a multi-stage estimation routine:

• Batch Differential Correction: Takes a sequence of bearing measurements gathered across an orbital pass and solves an initial non-linear least-squares optimization to generate a coarse state estimate for both the observer and the observed targets.
• Sequential Kalman Filtering: Ingests subsequent optical tracks via an Extended Kalman Filter (EKF) or Unscented Kalman Filter (UKF). The filter propagates state vectors and covariance matrices using high-fidelity numerical integrators that incorporate Earth gravity field harmonics ($J_2$ through $J_{10}$), solar radiation pressure, and atmospheric density models.
• Simultaneous Orbit Refinement: When multiple space debris objects are sighted within a single orbital revolution, the geometric intersections lock down the host spacecraft's state vector while isolating systematic errors in the cataloged debris ephemerides, updating both simultaneously.

This is the orbital mechanics equivalent of Simultaneous Localization and Mapping (SLAM), a foundational robotic technique re-engineered for hypervelocity space environments.

The Vulnerability of Spaceborne GPS and the Necessity of Alternatives

The push to navigate via space debris stems from structural vulnerabilities in relying on terrestrial Global Navigation Satellite Systems (GNSS) for space operations.

+-------------------------------------------------------------------------+
|                  VULNERABILITY MATRIX: GNSS IN SPACE                    |
+-------------------------------------------------------------------------+
| Threat / Limitation | Low Earth Orbit (LEO)  | Cislunar / High Orbits   |
+---------------------+------------------------+--------------------------+
| RF Jamming/Spoofing | Severe risk in warfare | Low direct risk          |
| Kinetic ASAT Target | High systemic risk     | Low direct risk          |
| Geometric Dilution  | Excellent coverage     | Catastrophic (past GEO)  |
| Signal Attenuation  | Nominal (-160 dBW)     | Weak sidelobes (-180 dBW)|
| Power / SWaP Cost   | Moderate dedicated load| High receiver complexity |
+---------------------+------------------------+--------------------------+

Civilian and military satellites in LEO have long depended on GPS (United States), GLONASS (Russia), Galileo (Europe), and BeiDou (China) for precise position, navigation, and timing (PNT). However, several key vulnerabilities make this single point of failure a growing concern:

1. Electronic Warfare and RF Vulnerabilities

GPS signals broadcast from medium Earth orbit (~20,200 km altitude) reach receivers on Earth and in LEO at power levels as low as $-160\text{ dBW}$—equivalent to a fraction of a quadrillionth of a watt. Terrestrial military conflicts have highlighted the ease with which ground-based and airborne electronic warfare systems can jam, blind, or spoof civilian and military GPS channels.

As counter-space capabilities expand, dedicated ground-to-space and space-to-space RF jammers can degrade GPS reception across wide orbital sectors. A satellite that relies exclusively on GNSS signals risks losing orbit determination capability during jamming events, potentially causing drift, lost pointing references, and collision hazards. Developing reliable navigation without GPS has transitioned from an academic interest into an operational priority for military and civilian space assets alike.

2. The Geometrical Limit: The Cislunar and Deep Space Drop-off

GPS antennas are engineered primarily to point downward toward the surface of the Earth. Satellites orbiting below the GPS constellation in LEO intercept the main transmission lobes without difficulty. However, once a spacecraft travels into High Earth Orbit (HEO), Geostationary Transfer Orbit (GTO), or beyond Geostationary Earth Orbit (GEO) into Cislunar space, it leaves the main broadcast cone entirely.

                         [ GPS Satellite ] (20,200 km)
                                |
                                | (Main Transmission Beam)
                                v
                           [  EARTH  ]
                                
                      . - ~ ~ ~ - .
                  .                   .   (Faint Sidelobe Signals)
                .                       .
               .                         .
     [ High Altitude Spacecraft / Lunar Probe ] ---> Operates outside main cone;
                                                     requires alternative navigation.

High-altitude spacecraft must rely on faint "spillover" signals known as sidelobes—transmissions that bypass Earth's limb. At lunar distances (~384,400 km), sidelobe signal strength drops significantly, requiring large high-gain antennas, highly sensitive cryo-cooled or advanced receivers, and prolonged signal integration times. In the lunar environment and deep space, GNSS-based navigation becomes impractical.

3. Space Traffic Management and Infrastructure Resilience

Even in LEO, the global space architecture faces physical threats. If a kinetic Anti-Satellite (ASAT) missile test, intentional attack, or cascading series of accidental collisions were to degrade a GNSS constellation in medium Earth orbit, hundreds of operational satellites could lose positioning simultaneously.

A decentralized spacecraft that relies on its own optical vision and an internal catalog can maintain precise orbital state estimation independently of any functional constellation or terrestrial control uplink.

Architectural Breakdown: The FALCON Algorithmic Engine

The operational success of the FALCON trial rests on solving three interlinked computing bottlenecks: dynamic feature extraction, the "lost-in-space" data association problem, and co-optimization filtering on resource-constrained flight hardware.

+---------------------------------------------------------------------------+
|               FALCON ONBOARD COMPUTATIONAL SUBSYSTEMS                     |
+---------------------------------------------------------------------------+
|                                                                           |
|  +------------------------+  +---------------------+  +----------------+  |
|  | Sighting & Extraction  |  | Data Association    |  | Dual-State EKF |  |
|  | - Star De-blending     |->| - Ephemeris Pruning |->| - Orbit Prop.  |  |
|  | - Centroiding          |  | - Hypotheses Match  |  | - Covariance   |  |
|  | - Streak Detection     |  | - Mahalanobis Gate  |  | - Co-Update    |  |
|  +------------------------+  +---------------------+  +----------------+  |
|                                                                           |
+---------------------------------------------------------------------------+

Module 1: Sighting, Extraction, and Sub-Pixel Centroiding

Star trackers capture continuous exposures of the celestial field. In standard operations, the exposure duration and sensor gain are tuned to capture point-source starlight while suppressing high-frequency transient signals.

FALCON reconfigures this sensor pipeline:

• Starfield De-blending: Point sources that match static star patterns within the onboard Hipparcos/Tycho sub-catalogs are identified. These stars establish the instantaneous attitude matrix $\mathbf{T}_{b}^{i}$ converting coordinates from the camera body frame ($b$) to the Geocentric Celestial Reference Frame ($i$).
• Streak & Point Centroiding: Resident space objects appear either as localized points moving with distinct apparent velocities or, in longer exposures, as faint streaks.
• Sub-Pixel Estimation: Using modified Gaussian point-spread function (PSF) fitting or Radon transform line extraction, the software determines the precise centroid coordinate $(u_p, v_p)$ of the debris image plane down to a fraction of a pixel. This yields an angular measurement precision often exceeding a few arcseconds.

       Pixel Grid Detection:
       +---+---+---+---+---+
       |   |   | * |   |   |   Gaussian PSF Peak Fitting
       +---+---+---+---+---+   ------------------------> Sub-pixel line-of-sight
       |   | * | X | * |   |   Calculates centroid to within fraction of a pixel.
       +---+---+---+---+---+
       |   |   | * |   |   |
       +---+---+---+---+---+

Module 2: The Lost-in-Space Data Association Engine

Once an optical streak is detected, the flight system faces an identification challenge: Which of the roughly 20,000 objects in the onboard catalog did the camera just observe?

                         [ Observed Optical Streak ]
                                     |
                                     v
                       +---------------------------+
                       | Spatial / Temporal Gating |
                       |    (Field of View Cone)   |
                       +---------------------------+
                                     |
                +--------------------+--------------------+
                |                                         |
                v                                         v
     [ Candidate Object A ]                    [ Candidate Object B ]
     (Within 3-sigma gate)                     (Outside 3-sigma gate -> Rejected)
                |
                v
     [ Mahalanobis Distance Matching ]
     d_M^2 = (z - h(x))^T S^(-1) (z - h(x))
                |
                v
     [ Highest-Likelihood Object Identified ]

The spacecraft may have experienced hours or days of accumulated unmodeled drift (the "lost-in-space" condition). If the spacecraft’s prior position estimate is uncertain by 50 kilometers, searching the entire 20,000-object database creates massive computational overhead and risks false-positive associations.

To solve this in real time on a low-power CubeSat processor, the algorithm executes structured catalog pruning:

1. Geometric Cone Gating: The software uses the camera's pointing boresight vector to filter the 20,000-object database down to objects whose predicted orbits intersect the sensor’s current field-of-view cone at time $t_k$.
2. Apparent Angular Velocity Filtering: The software compares the measured angular velocity vector ($\dot{\theta}_x, \dot{\theta}_y$) against the candidate targets. Orbiting debris objects exhibit angular rates directly tied to relative orbital inclination and altitude differences, immediately ruling out candidates with incompatible trajectories.
3. Statistical Hypothesis Testing: Remaining candidates are evaluated using a Mahalanobis distance metric:

$$d_M^2 = (\mathbf{z}_k - \mathbf{h}(\hat{\mathbf{x}}_k^-))^T \mathbf{S}_k^{-1} (\mathbf{z}_k - \mathbf{h}(\hat{\mathbf{x}}_k^-))$$

where $\mathbf{z}_k$ is the optical measurement, $\mathbf{h}(\hat{\mathbf{x}}_k^-)$ is the non-linear measurement model evaluated at the predicted state, and $\mathbf{S}_k$ is the innovation covariance matrix. Only observations passing a strict statistical threshold are admitted into the state estimator.

Module 3: Dual-State Co-Optimization Filter

The state vector tracked by the onboard filter is expanded beyond the host spacecraft to include the dynamic parameters of the observed debris targets:

$$\mathbf{X} = \begin{bmatrix} \mathbf{r}_{\text{host}} \\ \mathbf{v}_{\text{host}} \\ \mathbf{p}_{\text{host}} \\ \mathbf{r}_{\text{tgt}, 1} \\ \mathbf{v}_{\text{tgt}, 1} \\ \vdots \\ \mathbf{r}_{\text{tgt}, N} \\ \mathbf{v}_{\text{tgt}, N} \end{bmatrix}$$

where $\mathbf{p}_{\text{host}}$ represents empirical parameters such as camera alignment biases, sensor scale factors, and atmospheric drag coefficients ($C_D$).

+--------------------------------------------------------------------------+
|                  CO-OPTIMIZATION STATE VECTOR INTERACTION                |
+--------------------------------------------------------------------------+
|                                                                          |
|   +--------------------------+          +-----------------------------+  |
|   | Host Spacecraft State    |          | Debris Target States (1..N) |  |
|   |  r_host, v_host, Params  |          |   r_tgt, v_tgt, Ballistics  |  |
|   +--------------------------+          +-----------------------------+  |
|                \                              /                          |
|                 \                            /                           |
|                  v                          v                            |
|                 +----------------------------+                           |
|                 | Error Covariance Matrix P  |                           |
|                 | Cross-Coupling Terms P_ht  |                           |
|                 +----------------------------+                           |
|                                |                                         |
|                                v                                         |
|    Measurements update both Host State AND Target Ephemerides            |
|                                                                          |
+--------------------------------------------------------------------------+

When an optical measurement of debris object $i$ is processed, the Kalman gain matrix updates both the host spacecraft's state and the target’s orbital parameters through non-zero cross-covariance terms:

$$\mathbf{P} = \begin{bmatrix} \mathbf{P}_{\text{host}} & \mathbf{P}_{\text{host},\text{tgt}} \\ \mathbf{P}_{\text{host},\text{tgt}}^T & \mathbf{P}_{\text{tgt}} \end{bmatrix}$$

This dual-state estimation explains why FALCON was able to update and refine the known orbits of over 200 space debris objects during its three-day operational trial. The host satellite used the debris to locate itself; simultaneously, the collection of sightings allowed the spacecraft to compute higher-fidelity orbit solutions for the debris than the ground station tracking data held in its static catalog.

The Paradigm Inversion: Transforming Space Debris from a Threat into an Asset

The conceptual breakthrough demonstrated by Starling is not merely algorithmic; it represents an inversion of space environmental doctrine.

+-------------------------------------------------------------------------+
|                  THE DUAL PARADIGMS OF ORBITAL DEBRIS                   |
+-------------------------------------------------------------------------+
| Attribute          | Legacy Paradigm        | FALCON Paradigm           |
+--------------------+------------------------+---------------------------+
| Core Perception    | Threat / Collision Risk| Dynamic Navigation Beacon |
| High Density Value | Increased Danger       | Faster Fixes / Accuracy   |
| Tracking Objective | Evasive Maneuvers      | State-Vector Estimation   |
| Ground Role        | Sole Source of Truth   | Secondary / Peer Observer |
| Hardware Demands   | Complex Radar Shields  | Repurposed Star Trackers  |
+--------------------+------------------------+---------------------------+

The Historic Debris Burden

Since the launch of Sputnik in 1957, human spaceflight has generated a dense shell of orbital debris. The United States Space Surveillance Network (SSN), operated by the U.S. Space Force's 18th Space Defense Squadron, currently tracks roughly 36,500 individual objects larger than 10 centimeters. Models estimate that between 1 and 10 centimeters, over one million fragments drift through orbit untracked or cataloged only statistically.

                ESTIMATED ORBITAL DEBRIS POPULATION (LEO-GEO)
                +-------------------------------------------+
                | > 10 cm      | ~36,500 objects (Tracked)  |
                | 1 cm - 10 cm | ~1,100,000 fragments       |
                | < 1 cm       | > 100,000,000 particles    |
                +-------------------------------------------+

Historically, this population has been viewed entirely through the lens of risk:

• Kessler Syndrome: The risk that collisions between large objects generate cascading clouds of debris, rendering specific orbital planes unusable for generations.
• Conjunction Assessment Overhead: Ground control teams spend thousands of hours weekly analyzing Conjunction Data Messages (CDMs), often commanding costly propellant burns to dodge debris.
• Shielding Penalties: Extra mass allocated to Whipple shields and armored hulls to protect crewed and high-value payloads.

The Operational Utility of Debris

The FALCON experiment fundamentally alters this perspective. In the context of autonomous celestial tracking, every piece of orbital junk larger than a few centimeters is an illuminated reference beacon traveling along a mathematically predictable trajectory.

                         [ SUN ] (Broadband Illumination)
                            |
                            v
                  [ Orbital Debris Object ]
                   (Reflects Solar Photons)
                            |
                            v (Scattered Light)
    [ Star Tracker Camera on Autonomous Spacecraft ]
                            |
                            v
     (Resolves Line-of-Sight Bearing Angles)
                            |
                            v
     [ Onboard PNT Solution without GPS or Uplinks ]

This transforms orbital congestion from a purely negative hazard into a functional utility:

1. Density Enhances Geometry: In sparse orbital regimes, a spacecraft might wait hours between sightings of cataloged targets. In dense LEO bands (such as 500–900 km altitude), dozens of cataloged debris objects cross the field-of-view of a star tracker every single orbit. High debris density provides more frequent measurement updates, improving the precision of the onboard Kalman filter.
2. Autonomous Space Domain Awareness (SDA): Ground-based radar and optical telescopes face severe geometric and operational bottlenecks: atmospheric absorption, weather dropouts, geographic limits on sensor placement, and low revisit rates. When active spacecraft cross-reference and update debris orbits directly in space, the tracking burden shifts from overloaded ground infrastructure to decentralized orbital networks.
3. Closing the Space Traffic Loop: When an operational spacecraft spots debris whose trajectory differs from the onboard catalog, it refines the target's orbital parameters immediately. If shared over inter-satellite communications crosslinks, this updated ephemeris enables autonomous swarm-wide collision avoidance maneuvers without waiting for ground control commands.

Comparative Analysis: Alternative Modalities for GPS-Independent Space Navigation

To place the Starling FALCON experiment in context, it is instructive to compare optical debris tracking with other methods developed to provide navigation without GPS across varied spaceflight regimes.

+---------------------------------------------------------------------------------------------+
|                               COMPARISON OF PNT METHODS                                     |
+-----------------------+-------------------+--------------------+----------------------------+
| Navigation Technology | Primary Sensors   | Positional Accuracy| Operational Limitations    |
+-----------------------+-------------------+--------------------+----------------------------+
| Optical Debris (FALCON)| Repurposed 2D COTS| High (Tens to      | Requires catalog prior;    |
|                       | Star Trackers     | hundreds of meters)| dependent on solar lighting|
+-----------------------+-------------------+--------------------+----------------------------+
| X-ray Pulsars (XNAV / | Specialized X-ray | High (1-5 km in    | High mass, power, and cost;|
| SEXTANT)              | Timing Telescopes | deep space)        | long integration times     |
+-----------------------+-------------------+--------------------+----------------------------+
| Terrain Relative Nav  | Optical Cameras / | Ultra-High         | Only functions near surface|
| (TRN / LVS)           | LiDAR             | (Centimeters/Mtrs) | of Moon, Mars, or Asteroid |
+-----------------------+-------------------+--------------------+----------------------------+
| Intersatellite Cross- | RF Transceivers / | Ultra-High         | Requires cooperative nodes;|
| link Ranging (ISLs)   | Optical Terminals | (Centimeters/Mtrs) | shared constellation clock |
+-----------------------+-------------------+--------------------+----------------------------+
| Geomagnetic Field /   | Magnetometers /   | Low                | Prone to magnetic storms;  |
| Horizon Sensors       | Horizon Scanners  | (Several km)       | degrades rapidly with alt. |
+-----------------------+-------------------+--------------------+----------------------------+

1. X-ray Pulsar Navigation (XNAV / SEXTANT)

Demonstrated by NASA on the International Space Station via the NICER/SEXTANT payload, XNAV measures the periodic millisecond X-ray pulses emitted by rotating neutron stars (pulsars). Because pulsars have rotational stabilities comparable to atomic clocks, triangulating pulse arrival times from three or more pulsars provides an absolute navigation fix anywhere in the solar system.

+--------------------------------------------------------------------+
| Pulsar Timing Triangulation (XNAV):                                |
| [ Pulsar A ] ----\                                                 |
| [ Pulsar B ] -----> [ X-ray Detector / Atomic Clock ] => State Fix |
| [ Pulsar C ] ----/                                                 |
| * Advantage: Truly universal; works anywhere in the galaxy.        |
| * Disadvantage: Heavy X-ray optics; tens of watts of power;       |
|                 hours of integration time needed per fix.          |
+--------------------------------------------------------------------+

2. Terrain Relative Navigation (TRN)

3. Cooperative Intersatellite Crosslink Ranging

Mega-constellations like Starlink and government architectures like the Proliferated Warfighter Space Architecture (PWSA) utilize RF or laser crosslinks (optical inter-satellite links, or OISLs) to measure time-of-flight distances between constellation nodes.

FALCON, by contrast, operates completely passively by observing inert, non-cooperative matter.

Architectural Principles Extracted from the Case Study

Examining the operational success of FALCON allows us to extract four generalizable systems engineering principles applicable to broader aerospace, autonomous robotics, and defense architectures.

+-------------------------------------------------------------------------+
|                  CORE SYSTEMS ENGINEERING PRINCIPLES                    |
+-------------------------------------------------------------------------+
|                                                                         |
|  1. SENSOR DUAL-USE         Repurpose existing attitude sensors to      |
|                             eliminate dedicated navigation payload mass.|
|                                                                         |
|  2. EDGE CO-OPTIMIZATION    Estimate observer and environment states    |
|                             simultaneously to outpace ground tracking.  |
|                                                                         |
|  3. EXPLOITING RESIDUE      Convert operational clutter (debris) into   |
|                             the core infrastructure for positioning.    |
|                                                                         |
|  4. DECENTRALIZED SURVIVAL  Decouple mission-critical PNT functions     |
|                             from external RF and ground uplinks.        |
|                                                                         |
+-------------------------------------------------------------------------+

Principle 1: Sensor Dual-Use and SWaP-C Optimization

In aerospace design, Size, Weight, Power, and Cost (SWaP-C) dictate mission viability. Adding a dedicated navigation payload—such as an extra radar altimeter, deep-space optical telescope, or specialized GNSS receiver—incurs significant mass, power, and thermal penalties.

The Starling demonstration emphasizes the value of functional repurposing: exploiting the unused sensor capacity of star trackers that are already present on nearly every three-axis stabilized spacecraft. By routing existing sensor data streams to an advanced software processing engine, complex navigation capabilities can be integrated purely through flight software updates without altering spacecraft hardware.

Principle 2: Decentralized Edge Computing vs. Ground-in-the-Loop Latency

The traditional paradigm of orbital operations relies heavily on ground stations:

1. Space tracking radars ping an object.
2. Measurements flow to a central operations center.
3. Batch orbit determination is computed on terrestrial servers.
4. An updated ephemeris (e.g., Two-Line Element sets) is uplinked to the spacecraft during ground station passes.

Traditional Ground Loop:
[ Debris ] ===> [ Ground Radar ] ===> [ Control Center ] ===> [ Uplink Pass ] ===> [ Spacecraft ]
                 (High Latency, Intermittent Visibility, Vulnerable Comms)

FALCON Edge Architecture:
[ Debris ] ===> [ Repurposed Star Tracker ] ===> [ Onboard Processor (Era-Core) ]
                 (Zero Latency, Continuous Tracking, Fully Autonomous)

This cycle introduces hours or days of latency. Atmospheric density fluctuations can cause ground-predicted satellite positions to drift by kilometers over 48 hours. By moving the orbit determination engine directly onto edge flight processors, the Starling spacecraft updated and improved debris orbit paths autonomously in real time. The space architecture shifts from centralized, ground-dependent tracking to an autonomous, distributed sensing network.

Principle 3: Symbiotic Estimation (Mapping by Navigating, Navigating by Mapping)

FALCON illustrates the power of coupled state estimation. The host spacecraft cannot refine its own position without observing the debris, yet it cannot refine the debris's trajectory without tracking its own motion.

By framing orbit determination as a simultaneous dual-state optimization problem, both state estimates converge toward ground truth with every observation pass. In robotic autonomy, this eliminates the requirement for a static, pre-calibrated environment; the system refines its environmental map as an automatic byproduct of navigating through it.

Principle 4: Resilience Through Environmental Exploitation

Autonomous systems operating in extreme environments should not depend entirely on artificial support networks that can fail or be denied. Instead, robust systems exploit physical, unalterable properties of their operational environment.

Just as autonomous underwater vehicles (AUVs) navigate by matching benthic terrain gravity signatures or acoustic sea-floor features when radio communications cannot penetrate seawater, spacecraft can exploit the thousands of physical, free-flying kinetic bodies sharing their orbital shell. The orbital environment itself provides the reference frame.

Strategic and Tactical Implications for Defense and Space Domain Awareness

While NASA’s Starling mission is a civilian science and technology demonstration, the strategic and defense implications of autonomous optical navigation using space debris are significant.

+---------------------------------------------------------------------------+
|               TACTICAL DEFENSE IMPACT: SILENT RUNNING (EMCON)             |
+---------------------------------------------------------------------------+
|                                                                           |
|  Vulnerable Active Spacecraft:                                            |
|  [ Spacecraft ] ===== Radio Ranging / Uplinks =====> [ Ground Base ]      |
|         |                                                                 |
|         +---> Emits RF signatures (Detectable by hostile ELINT assets)    |
|         +---> Vulnerable to uplink jamming & RF spoofing                  |
|                                                                           |
|  Resilient Passive Spacecraft (FALCON):                                   |
|  [ Spacecraft ] <==== Passive Solar Reflection ===== [ Space Debris ]     |
|         |                                                                 |
|         +---> ZERO RF Emissions (Emission Control / Stealth)              |
|         +---> Completely immune to RF jamming & cyber uplink disruption   |
|                                                                           |
+---------------------------------------------------------------------------+

Emission Control (EMCON) and Stealth Operations

In military space operations, transmitting radio signals—whether downlinking telemetry, uplinking commands, or pinging active radars—instantly reveals a spacecraft's position, orbital plane, and operational status to adversary Electronic Intelligence (ELINT) and Space Domain Awareness systems.

A military satellite employing passive optical tracking operates in complete radio silence (Emission Control, or EMCON). It emits zero photons across the electromagnetic spectrum, accepting only ambient solar reflections from surrounding inert space debris. It calculates precise navigation without GPS, maintains formation with friendly assets, and maneuvers autonomously without providing electronic signatures that adversary sensors can detect or track.

+-------------------------------------------------------------------------+
|                  DEFENSE & SECURITY IMPLICATIONS MATRIX                 |
+-------------------------------------------------------------------------+
| Strategic Vector        | Tactical Reality                              |
+-------------------------+-----------------------------------------------+
| Contested Navigation    | Ensures orbital maneuvering despite full      |
|                         | GPS blackout or localized RF jamming.         |
+-------------------------+-----------------------------------------------+
| Responsive SDA          | Instantly detects unannounced maneuvers by    |
|                         | adversarial satellites during flybys.         |
+-------------------------+-----------------------------------------------+
| Tactical Proliferation  | Enables low-cost, expendable smallsat swarms  |
|                         | to operate with resilient autonomy.           |
+-------------------------+-----------------------------------------------+
| Autonomous Intercept    | Provides terminal guidance for rendezvous,    |
|                         | inspection, and active debris removal.        |
+-------------------------+-----------------------------------------------+

Counter-Space Jamming Resilience

In high-intensity conflicts, early salvos often involve localized or theater-wide GNSS denial. An adversary might employ high-power ground jammers or orbital jamming platforms to blind GPS reception over disputed regions.

Swarms of low-cost tactical satellites equipped with software like Era-Core can continue their intelligence, surveillance, and reconnaissance (ISR) orbits unimpeded. They track position by sighting adjacent space junk, calculate target ground coordinates, and maintain orbital phase geometry without referencing external positioning networks.

Autonomous Non-Cooperative Rendezvous & Inspection

The ability to identify, track, and compute orbits for non-cooperative objects from passive angles-only data is the fundamental algorithmic building block for rendezvous and proximity operations (RPO). Whether the objective is active debris removal, satellite servicing, inspection of an uncommunicative friendly asset, or monitoring a foreign spacecraft, this technology provides the trajectory estimation needed to guide intercept vectors autonomously.

Extensibility: Translating the Concept to Cislunar Space, the Moon, and Mars

The principles validated in low Earth orbit by the Starling FALCON experiment extend well beyond Earth-centered missions.

+---------------------------------------------------------------------------+
|                  EXTENSIBILITY TO CISLUNAR & DEEP SPACE                   |
+---------------------------------------------------------------------------+
|                                                                           |
|  LEO Environment (FALCON Base):                                           |
|  * Landmarks: Rocket bodies, dead satellites, fragments (~20,000 targets) |
|  * Gravity Field: Dominated by Earth J2-J10 harmonics                     |
|                                                                           |
|  Cislunar / Lunar Environment:                                            |
|  * Landmarks: Spent lunar transfer stages, historical orbiters (LRO),     |
|               Artemis infrastructure, lunar surface craters (Tycho)       |
|  * Gravity Field: Highly complex three-body dynamical regimes (NRHO, DRO) |
|                                                                           |
|  Interplanetary / Mars Environment:                                       |
|  * Landmarks: Natural asteroids (NEOs), Martian moons (Phobos, Deimos),   |
|               co-orbiting exploration spacecraft                          |
|  * Gravity Field: Solar / Heliocentric patched conics                     |
|                                                                           |
+---------------------------------------------------------------------------+

Navigating Cislunar Space and Lagrange Points

As NASA, international partners, and commercial entities advance the Artemis program, the volume of traffic operating in cislunar space—the expansive region between Earth and the Moon—is growing rapidly. Cislunar dynamics are governed by non-linear three-body circular restricted mechanics, where traditional Keplerian two-body approximations break down.

In cislunar orbits, such as Near Rectilinear Halo Orbits (NRHO) or Distant Retrograde Orbits (DRO), GPS signals are virtually non-existent. Spacecraft normally depend on time-constrained, expensive tracking support from NASA’s Deep Space Network (DSN) ground antennas.

Optical catalog navigation provides an autonomous alternative:

• Spacecraft can track spent translunar injection (TLI) booster stages, inactive lunar probes, communication relay satellites, and known near-Earth asteroids (NEAs).
• Optical cameras can simultaneously track distinct lunar landmarks—such as crater rims (e.g., Shackleton, Tycho, Copernicus)—and celestial debris.
• The onboard filter uses these sightings to determine exact orbital states within complex three-body manifolds, reducing reliance on DSN scheduling and freeing up ground antennas for scientific downlinks.

                             [ Earth-Moon L1 Point ]
                                       o
                                      / \
                                     /   \
                                    /     \
   [ Translunar Debris / Booster ] o       \
                                  /         o [ Inactive Lunar Probe ]
                                 /           \
                                v             v
                   [ Artemis Lunar Gateway / Swarm ]
                    (Passive Optical Cross-Sighting)

Deep Space Swarms and Interplanetary Visual Odometry

In future planetary exploration, distributed swarms of small spacecraft will be deployed to study asteroid belts, monitor Martian atmospheric dynamics, or explore the icy moons of Jupiter and Saturn.

Operating deep in interplanetary space with round-trip light communication delays ranging from several minutes to hours, these swarms cannot rely on real-time ground control. By tracking natural debris, small shepherd moons, parent asteroids, and neighboring swarm craft, multi-agent networks can autonomously maintain precise relative formations and compute heliocentric trajectories using visual odometry and angles-only filtering.

Technical Bottlenecks, Failure Modes, and Constraints

While the Starling FALCON trial demonstrated the viability of debris-based navigation, several physical and algorithmic challenges must be addressed before this technique can serve as a primary flight-critical navigation standard.

+-------------------------------------------------------------------------+
|                  FAILURE MODES & ENGINEERING BOTTLENECKS                |
+-------------------------------------------------------------------------+
| Bottleneck Category     | Physical Cause         | Mitigation Strategy  |
+-------------------------+------------------------+----------------------+
| Lighting & Phase Angle  | Eclipse / Umbra entry; | Integrate IMU / dead-|
| Constraints             | forward-scatter glare  | reckoning propagation|
+-------------------------+------------------------+----------------------+
| Ephemeris Degradation   | Atmospheric drag; solar| Ingest high-order    |
| & Orbit Drift           | storms; outgassing     | atmospheric models   |
+-------------------------+------------------------+----------------------+
| Edge Processor Overload | High-rate image PSF    | Prune spatial search |
|                         | fitting; 12-state EKF  | cones; GPU micro-ops |
+-------------------------+------------------------+----------------------+
| False Association /     | Dense debris clouds;   | Tight Mahalanobis    |
| Filter Divergence       | misidentified targets  | gating; multi-frame  |
+-------------------------+------------------------+----------------------+
+-------------------------------------------------------------------------+

1. Illumination Geometry and Solar Phase Angle Blind Spots

Optical cameras depend strictly on reflected sunlight. This creates two recurring operational blackouts during every orbit:

• Umbra / Eclipse Passes: When the satellite crosses into the Earth's shadow (approximately 35 minutes of a 90-minute LEO orbit), passing debris objects are not illuminated by the Sun. During this window, the star trackers can still see self-luminous background stars to maintain attitude, but non-luminous space debris is invisible. The navigation filter must propagate state estimates open-loop using dead-reckoning inertial models until orbital sunrise.
• Solar Glare / Low Phase Angles: Pointing an optical camera near the solar disc blinds the detector and can permanently damage CCD/CMOS arrays. If passing debris sits at high solar phase angles (scattering light away from the observer), its apparent visual magnitude drops below the star tracker's detection threshold (typically +8 to +10 magnitude for small lenses).

   [ Spacecraft in Umbra ] ----> Debris in Darkness (No Photons) ----> Filter Propagates via IMU
   [ Spacecraft in Sunlight ] -> Debris Reflects Photons (Visible) -> Filter Executes Kalman Update

2. High-Altitude Atmospheric Drag and Ephemeris Degradation

A catalog of 20,000 objects is not static; it decays continuously. In Low Earth Orbit below 600 kilometers, upper atmospheric density fluctuates dramatically based on solar activity, coronal mass ejections, and geomagnetic storms.