Research · Space Waste

RESEARCH · SPACE WASTE

Research

Published work, the open research agenda, and the engineering principles behind South Star PROX-M18, our RPODU (Rendezvous, Proximity Operations, Docking & Undocking) unit in development. We read the literature so the hardware doesn't have to learn it the hard way.

Published Reports

SW-TR-001 Rev B · Rendezvous, Proximity Operations, and Docking

An engineering assessment for active debris removal and on-orbit servicing, from approach through capture to takeover. NASA NESC lessons from DARPA Orbital Express, AR&D sensor heritage, attitude-sensing research, and post-capture takeover control. 14 pages.

DOWNLOAD PDF .DOCX

SW-TR-002 · The Last Ten Meters

Terminal-phase disturbances, plume impingement, and approach safety architecture. Built on NASA-JSC's CFD/DSMC plume modeling practice (EMPIRE lineage) and the ISS partnership's IRSIS corridor/abort standard. Its central inversion: the act of approaching disturbs the thing being approached. 7 pages.

DOWNLOAD PDF
“Nav is really important.” An Orbital Express principal designer, after the Scenario 3-1 recovery (NASA/TM-2011-217088). The thesis of SW-TR-001 in four words.
Research Agenda

Topics adjacent to the published reports. Every card links to a primary source. COVERED = treated in a published report. OPEN = targeted for future work.

OPEN · HIGH PRIORITY

Uncooperative RPO in practice: ADRAS-J

Astroscale's ADRAS-J (2024) rendezvoused with a real 3-ton H-2A upper stage using angles-only navigation, held 50 m fly-arounds, and closed to 15 m, the first proximity characterization of unprepared debris. The definitive validation case for SW-TR-001.

SOURCE: ASTROSCALE MISSION PAGE → OPEN · HIGH PRIORITY

Vision-based pose estimation & the domain gap

Stanford SLAB's SPEED/SPEED+ benchmarks expose the core ML problem: models trained on synthetic imagery degrade on real hardware-in-the-loop images (sunlamp glare, lightbox noise, Earth background). What spaceflight-qualified inference actually requires.

SOURCE: SPEED+ · ARXIV 2110.03101 → OPEN · HIGH PRIORITY

Net & tether-net capture dynamics

RemoveDEBRIS flew the first in-orbit net capture (2018); ESA's e.Deorbit program built the simulation methodology for nets wrapping rigid tumbling targets. Deployment parameters, capture quality indices, tether damping: the physics behind our net-capture pillar.

SOURCE: REMOVEDEBRIS · EOPORTAL → COVERED · SW-TR-002

The last ten meters: plume impingement & corridors

NASA-JSC's CFD/DSMC plume chain and flight-rule practice; IRSIS corridor, hold, and abort architecture; the debris inversion, where impingement torque perturbs the target's tumble exactly when margins are thinnest. Now a published report.

READ: SW-TR-002 (PDF) → OPEN

Detumble before capture

Contactless trade space: eddy-current braking (the ‘Eddy Brake’ studied for Ariane rocket bodies), propellant-free magnetic-pulse brake-and-traction, electrostatic detumble of GEO objects. Controllability of the chaser-target magnetic interaction is the open question.

SOURCE: ACTA ASTRONAUTICA → OPEN

Sensor ghosts near large reflective bodies

Laser and RF returns multipath off 11-meter metal cylinders: real return plus ghost reflections, the proximity-ops version of ground-bounce interference. Test methodology against reflective mockups; gating treatment sketched in SW-TR-002 §6.

RELATED: NTRS 20120017923 → OPEN

Program watch: open-source intelligence

China's Shijian-21 towed defunct Beidou-2 G2 past graveyard orbit (2022); SJ-21/SJ-25 performed the first GEO satellite-to-satellite refueling (July 2025). ClearSpace-1 and DARPA RSGS continue in parallel. A standing review of open OOS literature, always citing originals.

SOURCE: SPACENEWS → OPEN

Verification: predicted vs. measured

Air-bearing testbeds, hardware-in-the-loop campaigns, best-estimated trajectories (NESC R-2), and the prediction-vs-flight ledger kept from the first lab test onward. Credibility is the documented gap between model and measurement.

CONTEXT: SW-TR-001 §7 (PDF) → COVERED · SW-TR-001

Attitude sensing chain

Star-tracker/camera misalignment and active laser metrology; all-day star+Sun sensing from one aperture via oversaturated pixel response; magnetometer-free magnetic control. SW-TR-001 §3.

READ: SW-TR-001 (PDF) → COVERED · SW-TR-001

Post-capture takeover & inertia identification

Single-sample iterative inertia identification, Lyapunov takeover control, torque allocation; LiDAR-seeded estimation for debris; tether-tension identification. SW-TR-001 §8.

READ: SW-TR-001 (PDF) → OPEN · HIGH PRIORITY

Unintended emission in protected bands

A Curtin/ICRAR survey (Grigg, Tingay & Sokolowski, A&A 2025) logged 112,534 unintended emissions from 1,806 Starlink satellites, present in up to ~30% of low-frequency images and inside bands reserved for radio astronomy, outside the ITU's intentional-transmission rules. The measurement case for ground-based interference work.

SOURCE: A&A 699:A307 · ARXIV 2506.02831 → OPEN

Blind interference cancellation

Kratos SigX Protect cancels an interfering signal using equipment only on the receiving end, with no prior knowledge of the source, by characterizing it and generating a canceling signal in real time. The receive-side approach behind active interference cancellation.

SOURCE: KRATOS SIGX → OPEN

Distributed low-cost spectrum sensing

Electrosense+ and SatNOGS prove the pattern: inexpensive software-defined-radio nodes, on-device feature extraction, a shared backend and a public database, a planet-scale observatory built from dispersed antennas. The blueprint for a sensing minimum-viable network.

SOURCE: ELECTROSENSE+ · ARXIV 1811.12265 → OPEN

Machine learning for RF: detection & classification

Deep-learning spectrum sensing recovers signals that energy detection misses; CNN-based classifiers separate licit signals, flag spoofed ones, and detect signals in unknown, dynamic spectrum environments. The characterization engine, and a bridge to the orbital pose and guidance work.

SOURCE: DEEP LEARNING FOR RF · ARXIV 1909.11800 →
Field Notes · SW-RN-001 · Legacy Flight-Test Literature

Old aerospace and defense literature keeps solving our problems before we have them. A reading of declassified and archival NASA, MITRE, Army Research Laboratory, and CIA documents, mapped to where each applies in the Space Waste program.

Source What it established Where we apply it
MITRE flight-simulation study Declared its modeling assumptions up front (a local tangent-plane frame, valid over short ranges); aircraft tracking a planned route, altitude and speed profile; faster-than-real-time execution Approach-corridor guidance with a reference trajectory to correct against; accelerated-time simulation for AI training at scale
NASA linear-model derivations (1987–88) Derive the vehicle math once, rigorously; distribute as a shared software tool One canonical relative-dynamics library for the whole program, with no parallel re-derivations
Damaged-aircraft control study Controlling an aircraft whose mass properties change suddenly mid-flight Transfers almost directly to the moment of capture: sudden attachment of tumbling mass (SW-TR-001 §8)
Wind-shear landing studies The final seconds dominate the risk; model terminal disturbances explicitly The disturbance budget of SW-TR-002 §6: plume reflections and target wobble modeled, not margined
Minimum-time-climb study Solve the simplified problem first; use it to seed the full solution Warm-start trajectory optimizers and RL planners with the analytic answer instead of learning orbital mechanics from scratch
SR-71 performance validation Predict, fly, then explain every prediction-vs-flight discrepancy The predicted-vs-measured ledger, kept from first bench test onward
ARL ground-wave propagation Direct signal plus ground reflection interfere at the receiver Lidar/radar ghost returns near large reflective rocket bodies; mandates testing against reflective mockups
ARL beacon-navigation study Position and orientation from beacons mounted on the target The cooperative half of the business: where to advise satellite builders to mount fiducials and retroreflectors for serviceability
Artillery fire-control manual Baseline solution plus separately tabulated, individually tested corrections Guidance architecture: clean baseline plus auditable correction terms, reviewable piece by piece by defense customers (SW-TR-002 §6)
NASA airship study Don't claim superiority everywhere; define the missions where competitors structurally fail Positioning: small debris and tumbling derelicts are missions nobody else's architecture serves
CIA foreign-literature translations Reading what adversaries publish openly is the cheapest intelligence; every document traceable to an archive ID The Program Watch agenda item; citation discipline across all Space Waste publications
Operating Principles · SW-EN-001

Distilled from the flight-test and fire-control literature above. These are the habits we hold our own engineering to. SW-TR-002 opens with a declared-assumptions block, Principle 01 in practice.

01
State your assumptions first, and own themThe best flight-dynamics papers open by declaring their simplified world (a local tangent-plane frame, valid only over short ranges) and where it applies, without apology. Every Space Waste analysis leads with its model boundaries.
02
Derive once, share as a toolNASA derived its aircraft linear models once, carefully, and shipped them as software the whole team used. One source of truth for the core math; no parallel re-derivations drifting apart.
03
Three model tiers, always labeledA simple model for intuition, a medium model for design, a full-detail model for verification, and every result tagged with which tier produced it. NASA-JSC's plume practice runs exactly this way.
04
Spend your worry on the last ten metersWind-shear research concentrated on final approach because that is where landings fail. Docking is the same. SW-TR-002 exists because of this principle.
05
Keep the predicted-vs-measured ledgerThe SR-71 program predicted performance, flew, and explained every discrepancy. That ledger is what credibility is. Ours starts at the first bench test.
06
Small tested corrections beat one perfect equationArtillery fire control stacks a baseline solution with individually validated correction terms. The SW-TR-002 disturbance budget follows the same architecture, auditable piece by piece.
07
Label frames obsessively; cite originalsHalf of all integration failures are coordinate-frame mistakes; half of all credibility failures are unsourced claims. Every vector carries its frame, every claim carries its archive ID.
Reading List · Send Us Papers

The report series grows with its source library. Documents we are actively seeking:

  • ADRAS-J mission results and GN&C papers (Astroscale/JAXA CRD2, 2024–25): angles-only navigation, fly-around design, 15 m approach
  • SPEED+ / SPEC challenge papers (Stanford SLAB; arXiv 2110.03101) and follow-on pose-estimation work across the synthetic-to-real domain gap
  • RemoveDEBRIS flight results (net + harpoon experiments, 2018–19) and e.Deorbit net-dynamics simulation studies
  • Plume impingement for proximity ops: NASA EMPIRE high-fidelity simulations (NTRS 20120017923), DSMC thruster-plume methods, ISS visiting-vehicle plume requirements; IRSIS rendezvous interoperability standard
  • Contactless detumble: ‘Eddy Brake’ control analyses (Agora / Ariane rocket-body studies), magnetic-pulse brake-and-traction, electrostatic detumble of GEO objects
  • Air-bearing / HIL RPOD testbed methodology papers; best-estimated-trajectory reconstruction practice
  • Clohessy-Wiltshire / relative-motion guidance under uncertainty; passively safe trajectory design
  • Damaged-aircraft / control-reconfiguration literature after sudden mass-property change: directly transferable to post-capture takeover
  • Open literature on Shijian-21 / SJ-25 GEO operations, ClearSpace-1, DARPA RSGS, for the Program Watch file
  • DSMC / plume-target interaction studies for lightweight tumbling bodies: the SW-TR-002 §5 gap
  • RF geolocation by time-difference-of-arrival (HawkEye 360 RFGeo methodology and accuracy studies); adaptive-array and spatial-nulling references
  • Kratos SigX Protect technical notes; blind-separation and receive-side cancellation methods
  • Electrosense / SatNOGS network-architecture papers: distributed low-cost SDR sensing, shared backends, public databases
  • RF machine learning: automatic modulation classification, RFI excision, GAN-based transmitter identification, cyclostationary plus neural-network jammer detection
  • Starlink unintended-emission follow-on measurements affecting SKA-Low and LOFAR; ITU regulatory discussion of unintended emission