ADSC is not a debris collector. It is a low-cost concept for Kessler-precursor removal: a small servicer installs a passive deorbit kit on high-risk massive derelicts and departs, preventing future fragment clouds before they form — maximizing cleanup value per unit cost with a passively-safe approach design.
What this is: an open, reproducible evidence package for installer-type active debris removal — a C++17 GNC simulator plus generated, claim-audited artifacts. The product is evidence/adsc_evidence_pack.md: every number in it (and in this README) is machine-read from committed artifacts and regenerates byte-for-byte from a clean clone. This is not flight software and not a mission proposal (TRL 4, GNC software element, simulation environment).
Why: the collisional cascade's fuel is the population of massive derelict upper stages in congested bands — fragments are the symptom. The servicer targets the objects that would become the next fragment clouds, attaches a passive deorbit kit (drag sail or electrodynamic tether), and departs. What an agency already has — funded capture/tug programs (ClearSpace-1, ADRAS-J2) and published kit physics — does not include an open, honesty-audited trade for the installer architecture; that is the niche this package fills (docs/concept.md).
The placeholder-independent core: an installer never carries per-target deorbit Δv and never tows or detumbles a multi-tonne stage; batch amortization drives cost-per-removal to 28.5 % of the single-target baseline at N = 4 kits (Δv-limited, measured — see Key results). This is a mass-ratio scaling argument that survives every open parameter.
Closed-loop GNC (sliding-mode attitude control, translation EKF + attitude MEKF, guided approach with reachability-screened aborts), campaign Monte-Carlo (N = 500/catalog, Wilson CIs), a WP12 fidelity ladder (CW → +J2 → +drag; estimate-driven guidance; minimum-impulse-bit actuator), kit decay trades (sail + WP13 EDT physics band), a relative-unit cost model with WP14 absolute-cost ranges, a regulatory precheck (not legal advice), and the generated evidence pack. Module-by-module detail: docs/technical_architecture.md · docs/gnc.md.
Deliberately not implemented: flight-qualified code (WP9 reserved — the only path to TRL 5), inertia identification, plane-change optimization, homebrew breakup models, IRI plasma, absolute point-value costs, legal determinations. Full honest scope: docs/limitations.md and the evidence pack's PLACEHOLDER inventory (140 marks, 56 decision-critical — one glance away, by design).
| campaign (N = 500/catalog) | SL-16 class | SL-8 class |
|---|---|---|
| productive-end rate | 0.556 [0.512, 0.599] | 0.542 [0.498, 0.585] |
| keep-out violations | 0.000 [0.000, 0.008] [L0, ds-v1] | 0.000 [0.000, 0.008] [L0, ds-v1] |
| Δv used p50 [m/s] | 124 | 124 |
| removals/mission p50 | 4 | 4 |
Full tables + failure taxonomy: docs/safety.md.
| cost/FoM (relative CU; WP14 MUSD ranges) | value |
|---|---|
| amortization minimum (N = 4) | 44.80 CU/removal = 0.285× the N = 1 baseline |
| FoM p50, SL-16 (spatial / criticality) | 156.70 / 200.90 kg/CU |
| derived anchor (low / mid / high) | 0.0628 / 0.1546 / 0.4213 MUSD/CU |
| cost/removal p50 at the anchor | 2.81 / 6.92 / 18.87 MUSD |
Full model + sources-or-PLACEHOLDER itemization: docs/cost_model.md.
Kit-class trade (WP13): sail closes for the ~1.4 t class (7..115 m²) but not for the ~9 t class (135..2155 m²); EDT is an honest candidate with open risks for the heavy class (2.98..9.17 yr band; libration T7 unresolved) — docs/target_selection.md.
Approach safety is a passively-safe approach design with clearance-verified aborts (keep-out violations 0/500 per catalog [L0, ds-v1, Wilson ≤ 0.0076]), re-verified at L1 (+J2) and L2 (+drag) with zero violations over the 292/291 re-verified abort events [ds-v1/ds-v2] — never "strict approach safety": the guarantees are model-scoped (R14 tags everywhere), and a real mission would re-verify against a higher-fidelity propagator. docs/safety.md.
cmake -S . -B build -DCMAKE_BUILD_TYPE=Release && cmake --build build
bash tools/regenerate_all.sh build # regenerates EVERY committed artifact
CI enforces byte-identity of generated/, evidence/, this README's number
blocks, and docs/ on every push. Requires C++17 + Eigen 3.3; Python 3
stdlib only for tooling. Details: docs/technical_architecture.md.
evidence/ the product: generated, claim-audited evidence pack
docs/ concept, architecture, gnc, safety, targets, cost,
legal, limitations, roadmap, adoption brief, one-pager,
5-page summary, deck source, release engineering
generated/ committed machine-readable artifacts (CSV/md/SVG)
include/ src/ tests/ C++17 simulator + unit tests
tools/ stdlib-only generators, audits, compliance precheck
adsc-specification-v5.md the binding spec (rules R1-R16, decisions D1-D13)
WP1–WP8 ✅ · WP10 forensics ✅ · WP11 safety hardening ✅ · WP12 fidelity ladder ✅ · WP13 kit trade + EDT ✅ · WP14 cost ranges + FoM ✅ · WP15 proposal package ✅ (this release) · WP9 processor-in-the-loop — reserved, not started, the only path above TRL 4. Detail: docs/roadmap.md.
TRL 4 applies to the GNC software element only; system-level TRL is not claimed. Passive-safety numbers hold in the stated models, not the real environment. Decision-critical inputs remain PLACEHOLDER until cited (inventoried automatically in the evidence pack §10). docs/limitations.md is the honest one-stop list.
MIT (LICENSE). Conceptual prototype for peaceful ADR research and education only; no military or re-entry-weapon application is intended. Regulatory outputs are research prechecks, not legal advice.
本プロジェクトは平和的デブリ除去技術の教育・研究目的の概念プロトタイプです。 実際の宇宙機運用には専門機関による設計審査・認定・国際法遵守が必要です。