Deep Research (paper-location pass) — Arm singularity & base–EE null-space coupling

Date: Jun 9, 2026 · Pairs with: singularity_literature_brief.md, deep_research_Jun08.md (don’t repeat). Method: Semantic Scholar MCP, queried per the brief’s four targeted searches. All metadata (authors, year, venue, DOI, citation count, open-access status) is verified from Semantic Scholar.

STATUS — read this first. This is a paper-FINDING pass, not a full-text synthesis. Identities and download status are verified. The “answer-first” fork below is a preliminary framing built only from (a) abstracts I actually retrieved, (b) facts already established in deep_research_Jun08.md, and (c) paper titles/venues. Claims needing the full text are tagged [VERIFY ON FULL TEXT]. Nothing here invents a finding or an equation from a paper I have not read.

Answer-first: the redundancy-vs-managed-regime fork (preliminary)

The literature splits the brief’s strategic fork into three options, not two — and the cheapest one is under-explored in your stack:

Stay managed (speed-modulation + Tikhonov), but upgrade the inverse. Your §4.6 plain Tikhonov is the floor of singularity-robust IK. Chiaverini 1997 (filtered/selective DLS + task priority, 726 cites) and Nakamura–Hanafusa 1986 (the SR-inverse origin) damp only the near-singular direction instead of the whole solution — which should buy less p_e derating for the same robustness. This directly attacks your “p_e tracking tax” without new hardware. [VERIFY ON FULL TEXT] that selective damping applies cleanly to the circumcentroidal J⁺.
Null-space damping instead of redundancy (Q1). The Nenchev–Yoshida Reaction Null Space line is the literature’s tool for exactly the “indirectly damped 1-DOF arm mode” Giordano 2020 blamed for the wobble. If the wobble and the elbow singularity are one phenomenon (your key link), explicit RNS/null-space damping may suppress it without a 7th joint. Caveat the brief already flags and the abstracts corroborate: RNS paths are nonholonomic and singularity-rich — so this trades one singularity problem for another. [VERIFY ON FULL TEXT]
Go structural — 7-DOF redundancy. Genuinely thin literature for free-flyers specifically (see Q3): no canonical anchor, mostly 2025 one-offs. The closest grounded guidance is the Calzolari 2020 singularity-map / gradient-planning and Giordano 2018 workspace-fixation papers (your own DLR/Giordano group, both open access). A 7th joint also breaks the square-Jacobian circumcentroidal formulation — a real cost the brief named.

Preliminary lean: exhaust (1) before (3). The selective-DLS upgrade is low-cost, directly targets the tracking tax, and is well-anchored; redundancy is high-cost and under-supported in the free-flyer literature. (2) is the interesting wildcard precisely because it tests your “wobble = elbow singularity” hypothesis. This lean is from abstracts/titles only — confirm on full text before committing.

Located papers by question

Legend — Have?: ✅ already in Papers_and_Textbooks/ · ⬇️ open-access, downloadable now · 🔒 paywalled (abstract only).

Q1 — Reaction Null Space & null-space damping

Have?	Authors	Yr	Title	Venue	DOI / ID	Cites
🔒	Nenchev, Yoshida, Vichitkulsawat, Uchiyama	1999	Reaction null-space control of flexible structure mounted manipulator systems	IEEE T-RA	10.1109/70.817666	198
🔒	Nenchev, Yoshida, Uchiyama	1996	Reaction null-space based control of flexible structure mounted manipulator systems	IEEE CDC	10.1109/CDC.1996.577417	34
🔒	Nenchev, Yoshida, Vichitkulsawat, Konno, Uchiyama	1997	Experiments on RNS-based decoupled control of a flexible structure mounted manipulator	IEEE ICRA	10.1109/ROBOT.1997.619341	26
⬇️	Ye, Dong, Hong	2019	Adaptive Reaction Null Space Planning & Control (VFF–RLS & SSADE–ELM) for Free-Floating Space Robot	Electronics	10.3390/electronics8101111	3

Q2 — Dynamic vs kinematic singularities of the Generalized Jacobian

Have?	Authors	Yr	Title	Venue	DOI / ID	Cites
✅	Papadopoulos & Dubowsky	1993	Dynamic Singularities in Free-floating Space Manipulators	ASME JDSMC	10.1115/1.2897406	162
🔒	Papadopoulos & Dubowsky	1991	On the nature of control algorithms for free-floating space manipulators	IEEE T-RA	10.1109/70.105384	293
🔒	Papadopoulos & Dubowsky	1989	On the dynamic singularities in the control of free-floating space manipulators	(conf)	— (CorpusId 118837562)	81
🔒	Umetani & Yoshida	1989	Resolved motion rate control of space manipulators with GJM	IEEE T-RA	10.1109/70.34766	620
⬇️	Umetani & Yoshida	1989	(open-access JRSJ version of the GJM paper)	J. RSJ	10.7210/JRSJ.7.327	23
⬇️	Calzolari, Lampariello, Giordano	2020	Singularity Maps of Space Robots and their Application to Gradient-based Trajectory Planning	RSS	10.15607/rss.2020.xvi.015	7

Q3 — Kinematic redundancy (7-DOF) for singularity avoidance in free-flyers

Have?	Authors	Yr	Title	Venue	DOI / ID	Cites
⬇️	Giordano, Calzolari, Albu-Schäffer	2018	Workspace Fixation for Free-Floating Space Robot Operations	IEEE ICRA	10.1109/ICRA.2018.8460478	14
🔒	Chiaverini, Oriolo, Walker	2008	Kinematically Redundant Manipulators (Springer Handbook of Robotics)	Springer	10.1007/978-3-540-30301-5_12	199
🔒	Ma, Xie, Jiang et al.	2025	Robotic redundancy via arm-angle self-adaptation through nullspace resolution	IJRR	10.1177/02783649251371735	4
🔒	Liu, Li, Shi, Jin	2025	Real-time Trajectory Planning for Redundant 7-DOF Space Manipulators with Unknown Link Lengths	ICRCA	10.1109/ICRCA64997.2025.11011190	0

Finding: there is no canonical “redundancy resolves free-flyer singularities” anchor the way Chiaverini is canonical for fixed-base. The on-point material is recent, scattered, and low-citation. Treat the 7-DOF fork as less literature-supported than the managed-regime upgrade — a point in favor of option (1).

Q4 — Singularity-robust inverses beyond plain Tikhonov

Have?	Authors	Yr	Title	Venue	DOI / ID	Cites
🔒	Chiaverini	1997	Singularity-robust task-priority redundancy resolution for real-time kinematic control	IEEE T-RA	10.1109/70.585902	726
🔒	Nakamura & Hanafusa	1986	Inverse kinematic solutions with singularity robustness for robot manipulator control	ASME JDSMC	10.1115/1.3143764	1178
🔒	Bruschi & Invernizzi	2025	Singularity-free task-priority design for trajectory tracking in space robots	ACC	10.23919/ACC63710.2025.11107894	1

Supporting — manipulability & the “coupling-as-resource” counter-current

Have?	Authors	Yr	Title	Venue	DOI / ID	Cites
🔒	Yoshikawa	1985	Manipulability of Robotic Mechanisms	IJRR	10.1177/027836498500400201	2582
🔒	Yoshikawa	1985	Manipulability and redundancy control of robotic mechanisms	IEEE ICRA	10.1109/ROBOT.1985.1087283	319
⬇️	Das, Choi, Kim	2025	Understanding and Utilizing Dynamic Coupling in Free-Floating Space Manipulators for OOS	arXiv	2508.15732	0

Download shortlist (open access — grab these first)

Calzolari et al. 2020, Singularity Maps — RSS proceedings (free): roboticsproceedings.org/rss16/p015.pdf ⟵ highest priority; your group, Q2+Q3.
Giordano et al. 2018, Workspace Fixation — DLR: elib.dlr.de/126555/1/ICRA18_WorkspaceFixation_final.pdf
Das, Choi & Kim 2025, Dynamic Coupling — arXiv: arxiv.org/pdf/2508.15732
Ye et al. 2019, Adaptive RNS — MDPI: mdpi.com/2079-9292/8/10/1111/pdf
Umetani & Yoshida 1989 (GJM, JP open version) — J-STAGE: jstage.jst.go.jp/article/jrsj1983/7/4/7_4_327/_pdf

The high-value paywalled four (institutional pull / library)

Chiaverini 1997 (10.1109/70.585902) — the selective-DLS / task-priority backbone for option (1).
Nakamura–Hanafusa 1986 (10.1115/1.3143764) — SR-inverse origin.
Nenchev–Yoshida 1999 (10.1109/70.817666) — canonical RNS control, for option (2).
Bruschi–Invernizzi 2025 (10.23919/ACC63710.2025.11107894) — newest space-specific task-priority singularity handling.

Still to do / gaps

Full-text extraction per the brief’s template (singularity criterion / null-space treatment / soft-attitude applicability / proposed fix / reported cost) — pending; needs the PDFs above.
Not re-fetched this pass (already known from best_start.md, list for completeness): Wampler 1986 DLS (10.1109/TSMC.1986.289285); Chiaverini–Siciliano–Egeland 1994 DLS review (10.1109/87.294335). Both paywalled-likely.
Corroboration rule: load-bearing claims still need ≥2 sources once full texts are read (per the deep_research convention). This pass establishes which papers; it does not yet cross-validate their claims.