US 9566986

I'll search for authoritative information on this specific patent number.

I have the authoritative full text plus corroborating search results. Let me try to retrieve the verbatim claim set for this specific patent, since the provided full text cuts off before the claims.

Let me make one more targeted attempt to retrieve the verbatim granted claims of US9566986 (as opposed to its later continuations).

US Patent 9,566,986 — Analyst Summary

Source note: I relied on the authoritative full text fetched from Google Patents (https://patents.google.com/patent/US9566986/en) plus cross-checks against Unified Patents, uspto.report, Justia, and USPTO Patent Center/PTActs records. I did not auto-correct any identifier; all numbers below are literal.

---

1. Bibliographic data

Assignment chain (per Google Patents reassignment records):
IBM → DAEDALUS GROUP LLC (2019-11-14) → DAEDALUS GROUP, LLC (2020-01-27) → SLINGSHOT IOT LLC (2020-01-29) → GRANITE VEHICLE VENTURES LLC (2023-07-27).

Discrepancy flagged (not auto-corrected): Unified Patents lists the priority date as 2015-09-24, grant date 2017-02-13, and expiration 2035-09-24 — i.e., one day earlier than Google Patents' 2015-09-25 / 2017-02-14 / 2035-09-25. I am reporting both rather than resolving them; the Google Patents full-text record and the USPTO PTActs papers both date the IBM application to Sep. 25, 2015, which I regard as the more reliable figure, but the discrepancy is real and I cannot fully explain it.

---

2. Abstract (verbatim)

"A computer-implemented method, system, and/or computer program product controls a driving mode of a self-driving vehicle (SDV). Sensor readings describe a current operational anomaly of an SDV that is traveling on a roadway. One or more processors compare a control processor competence level of the on-board SDV control processor that autonomously controls the SDV to a human driver competence level of a human driver in controlling the SDV while the SDV experiences the current operational anomaly. One or more processors then selectively assign control of the SDV to the on-board SDV control processor or to the human driver while the SDV experiences the current operational anomaly based on which of the control processor competence level and the human driver competence level is relatively higher to the other."

---

3. Plain-language overview

The patent addresses a hand-off problem: when a self-driving vehicle develops a fault (bad brakes, bald tires, a burned-out headlight, icing, a software/firmware/hardware bug, etc.), which "driver" is better at handling that specific problem — the car's on-board computer or the human sitting in it?

The disclosed logic:

1. Sensors report an operational anomaly of the SDV (a fault or abnormal operating state).
2. The system computes a control processor competence level (CPCL) — how well the on-board SDV control processor handles this kind of anomaly, derived either from history (accident rate, speed, braking/starting, fuel economy for this or similar SDVs) or from a capability analysis (e.g., cruise control only vs. speed + following-distance + steering).
3. The system retrieves a driver profile giving a human driver competence level (HDCL) for the same anomaly type, again history-based or trait-based (e.g., a night-driving-restricted licence ⇒ low competence for a failed headlamp).
4. The two levels are compared — optionally factor-by-factor with weighted factors — and control is assigned to whichever is relatively higher (autonomous vs. manual mode).
5. The patent also covers: alerts before an automatic mode switch; handing the choice to the driver; cohort-based analysis (drivers or SDV processors sharing traits with a cohort whose accident data is known); weighted-and-summed sensor readings exceeding a threshold ⇒ manual mode prohibited; a fault-remediation table that maps fault → manifested condition → required driving mode, updated from a central server; and a minimum competence floor — if neither the processor nor the human is competent, the SDV autonomously pulls over and stops. It also describes rerouting to a resource provider capable of fixing the fault.

---

4. Independent claims — important caveat

I cannot give you verbatim independent-claim text for US 9,566,986 from an authoritative source, and I am flagging that rather than reconstructing it. The full-text record I was given is truncated in the detailed description, before the claims section, and my searches surfaced claim language only for later continuations of the same family, not for the '986 patent itself. So the following is a characterization grounded in the patent's own Summary/abstract, not a quote.

Based on the Summary, the '986 patent is a method / system / computer-program-product trio (the standard three statutory-class independent claims), each independently claiming:

• Receiving sensor readings describing a current operational anomaly of an SDV traveling on a roadway, the SDV being operable in autonomous mode by an on-board SDV control processor;
• a driving mode module that selectively controls autonomous vs. manual mode (manual = human driver controls the SDV);
• determining a control processor competence level of the on-board SDV control processor for that anomaly;
• receiving a driver profile giving a human driver competence level for that anomaly;
• comparing the two competence levels; and
• selectively assigning control of the SDV to the processor or the human based on which competence level is relatively higher.

If you need the exact granted claim language and claim count, pull the '986 patent's own claim set from USPTO Patent Center / Patent Public Search (or the granted-PDF images), not from a family member. Claim counts and claim scope in this family differ markedly between the '986 patent and its continuations.

---

5. Litigation check — please read this carefully

• Google Patents displays a "Family has litigation — First worldwide family litigation filed" flag with a Darts-IP link (patents.darts-ip.com/?family=57964734...) on the '986 record. That is a family-level indicator only; Google explicitly disclaims having performed any legal analysis.
• I did not find, and cannot confirm, any CAFC 2026 docket or appeal specifically involving US 9,566,986. My searches did not surface a Federal Circuit 2026 docket naming this patent number. I will not invent one.
• What I can confirm is that the family has been actively prosecuted and the assignee changed hands repeatedly (IBM → Daedalus → Slingshot IOT → Granite Vehicle Ventures), with continuation filings as recent as 18/743,841 (filed 2024-06-14, published as US 2025/0018963 A1). If you need a definitive docket answer, query the CAFC docket system and PACER/Docket Navigator directly by patent number — my tooling is not a substitute for that.

---

6. Family members — distinct numbers, do not conflate with 9,566,986

The '986 patent is the root of a continuation chain. These are different patents and their claims are not the '986 claims:

Related but separate IBM sibling disclosures with the same title include US 9,785,145 B2 (different abstract/claim scope) — also not the '986 patent.

---

7. Uncertainty statement

• High confidence: title, application number, inventors, original assignee, filing/priority date (2015-09-25), issue date (2017-02-14), abstract text, and the substance of the disclosure.
• Moderate confidence: current assignee chain, examiner name, and the one-day bibliographic discrepancy noted in §1.
• Low confidence / unverified: the verbatim granted independent claims and claim count of US 9,566,986; and the existence of any CAFC 2026 appeal naming this patent. I have explicitly declined to fabricate either.

I'll research litigation involving US patent 9566986. Let me run several searches across litigation-tracking sources.

Let me search litigation-tracking portals and current assignee entities more specifically.

The Tesla case appears to assert family members (US11597402, US11738765, US12037004). Let me verify whether US9566986 itself was ever asserted, and search for any other cases.

Summary of findings

No litigation asserting U.S. Patent No. 9,566,986 itself was found in the sources I was able to search. What exists is family-level litigation over sibling/continuation patents in the same family (same specification, same title "Controlling driving modes of self-driving vehicles," same inventors Gordon/Kozloski/Kundu/Malkin/Pickover). I want to flag this distinction clearly, because the strict rule here is to interpret identifiers literally — and 9,566,986 is not among the patents I can confirm as asserted.

---

Litigation found (asserts family members, not 9,566,986)

Granite Vehicle Ventures LLC v. Tesla, Inc.

Sources: E.D. Tex. docket (Case # 2:24-cv-01007-JRG, filed 12/06/2024) — https://dockets.justia.com/docket/texas/txedce/2:2024cv01007/[234675](/patent/234675) ; RPX Litigation — https://litigation.rpxcorp.com/litigation/txedce-234675-granite-vehicle-ventures-llc-v-tesla-inc ; Law360 case page — https://www.law360.com/cases/67536ec52ed14705899b498f/articles ; N.D. Cal. docket listing (3:2026cv01457) — https://dockets.justia.com/search?parties=Granite%20Vehicle%20Ventures%20LLC ; asserted-patent identification per IP-NAVI dispute record — https://www.ip-navi.or.kr/ipnavi/dispute/daily/dailyDetailPopup.navi?newsSeq=[218171](/patent/218171)

Related PTAB / appellate proceedings (same family, not '986)

• IPR2025-00943, -00944, -01034, -01035 — Tesla, Inc. v. Granite Vehicle Ventures LLC, filed May 5, 2025 (IPR2025-00943 challenges US 11,597,402; IPR2025-01034 challenges US 12,037,004). The Director denied institution on time-to-trial grounds (order dated Dec. 17, 2025; unsealed public version Dkt. No. 130-1).
• In re Tesla, Inc., No. 2026-116 (Fed. Cir. Feb. 27, 2026) — Federal Circuit denied Tesla's mandamus petition (nonprecedential), holding the institution-denial challenge unreviewable under Mylan v. Janssen. https://www.courtlistener.com/opinion/[10801275](/patent/10801275)/in-re-tesla-inc/
• Supreme Court — Tesla filed a petition for certiorari; an extension of time to file was sought (due July 27, 2026). Amicus briefs from CCIA, HTIA, Alliance for Automotive Innovation, National Retail Federation, etc. were filed in support. https://www.supremecourt.gov/DocketPDF/25/25A1289/... ; https://ipwatchdog.com/2026/08/31/computer-automotive-retail-industries-support-teslas-cert-petition-usptos-discretionary-[ipr](/resources/cost-of-inter-partes-review)-denials/
• IPR2025-01034 (US 12,037,004) — Patent Owner's discretionary-denial brief references the underlying district court case. https://ptacts.uspto.gov/ptacts/public-informations/petitions/[1557934](/patent/1557934)/...

---

On US 9,566,986 specifically

• The Google Patents page for US 9,566,986 carries a "Family has litigation" flag linking to Darts-IP family 57964734 (https://patents.darts-ip.com/?family=57964734...). That flag is family-level, and the underlying suits I could identify are the Granite v. Tesla actions over the family's continuation patents — not suits naming the '986 patent itself.
• The asserted patents recorded in the dispute databases for Granite v. Tesla are US 11,597,402, US 11,738,765, and US 12,037,004. US 9,566,986 does not appear among them.

Confidence and limitations (stated explicitly)

• High confidence: Granite Vehicle Ventures LLC v. Tesla, Inc., 2:24-cv-01007-JRG (E.D. Tex., filed 12/06/2024), transferred to N.D. Cal. as 3:26-cv-01457, asserting US 11,597,402 / 11,738,765 / 12,037,004; and the related PTAB/Federal Circuit/Supreme Court proceedings.
• Lower confidence / not fully verified: I could not locate any filed case in which US 9,566,986 is itself an asserted patent. My searches did not fully cover PACER directly, the Unified Patents litigation case list (the tool returned no separate case list for "9566986"), or Darts-IP's family record. It therefore remains possible — though I found no evidence — that the '986 patent was asserted in an earlier or separate action not surfaced here.
• Date caveat: The stated current date (April 26, 2026) appears stale — several retrieved sources are dated May–August 2026 (e.g., the Aug. 31, 2026 IPWatchdog amicus article and the May 18, 2026 Supreme Court filing). The status descriptions above reflect the most recent sources retrieved.

Bottom line: Based on the available evidence, there is no known litigation asserting U.S. Patent No. 9,566,986; the known litigation in this patent family is Granite Vehicle Ventures LLC v. Tesla, Inc. (E.D. Tex./N.D. Cal.) over the '402, '765, and '004 continuations.

I'll verify the structured data against live sources before drafting.

AIA Trial Proceedings — US Patent 9,566,986

Analysis date: 2026-10-01

Canonical source for this report is the structured PTAB proceedings on file block supplied with this task (USPTO Open Data Portal). Web search was used only to fill in and cross-check.

---

Proceedings overview

There are zero AIA trial proceedings on US Patent 9,566,986 (0 active / 0 claims invalidated / 0 claims sustained / 0 settled / 0 institution denied) — the ODP structured data returns no IPR, PGR, or CBM for this patent, and my web search found no petition, institution decision, FWD, or Board docket naming 9,566,986 as the challenged patent. The defensive posture is therefore "untested," not "hardened" and not "gutted": no claim of 9,566,986 has ever been canceled, and none has ever been upheld. The patent's entire family, however, is in active, high-profile litigation and has attracted five Tesla IPR petitions — all aimed at the continuation patents (11,597,402, 11,738,765, 12,037,004), each of which claims priority to 9,566,986. Those petitions were denied institution on discretionary grounds and are now at the Supreme Court — so there is still no claim-level PTAB ruling anywhere in this family. Flagging that explicitly because it is the single most important fact for a defendant: the "PTAB has already killed the claims" story does not exist here.

Important qualifier on the canonical data: ODP ingest lags. The Tesla petitions on the family were filed 2025-05 and later. If a petition against 9,566,986 were filed in the same window, it is possible ODP has not yet indexed it. My search surfaced three of Tesla's five petition numbers, and none is against 9,566,986 — but I could not verify the remaining two, so I cannot rule out the possibility at a level above "no evidence found."

---

Proceedings on US 9,566,986

None on file

• Type: N/A
• Filed: N/A
• Status: No AIA trial proceeding on file (per USPTO ODP structured data; confirmed by web search as of 2026-10-01)
• Judge panel: N/A
• Petition grounds: N/A — no petition against this patent
• Institution decision: N/A
• Final Written Decision: N/A — no claim of 9,566,986 has been canceled or sustained at the PTAB
• Settlement / termination: N/A
• Appeal: N/A as to this patent
• Defensive value: You cannot rely on any IPR estoppel, any FWD, or any prior Board claim construction for this patent. Anything you want to say about 9,566,986's validity at the PTAB, you have to file yourself.

Why this is unsurprising: 9,566,986 has never been asserted in the litigation I could identify. Granite Vehicle Ventures' complaint against Tesla (Granite Vehicle Ventures LLC v. Tesla, Inc., No. 2:24-cv-01007-JRG, E.D. Tex., filed 2024-12-06) and its P.R. 3-1 infringement contentions assert only claims of the three continuations — '402 claims 4, 6, 8, 9–11, 13; '765 claims 1, 3–8, 10–20; and '004 claims 1–24, 27, 29–39. The '986 is named in that litigation only as the common priority application ("Each asserted patent claims priority to U.S. App. No. 14/865,393, and each is therefore entitled to a priority date of at least September 25, 2015"). Petitions follow assertions; absent assertion, no petition.

Also worth noting for completeness: PGR is unavailable (nine-month post-grant window closed ~2017-11), and CBM would have been a non-starter for a self-driving-vehicle claims — the patent is not directed to a "financial product or service," and the CBM program sunset in 2020.

---

Adjacent AIA proceedings against the 9,566,986 family — contextual, NOT proceedings on 9,566,986

These are included only because the substantive art, the priority-date fight, and the estoppel analysis all carry over. Do not cite them as adjudications of 9,566,986.

IPR2025-00943 — Tesla, Inc. v. Granite Vehicle Ventures LLC (U.S. Patent 11,597,402)

• Type: Inter Partes Review
• Filed: 2025-05-05 (Patent Owner's mandatory notices dated 2025-05-22)
• Status: Institution denied (deputy/director-level discretionary denial on time-to-trial grounds); mandamus denied by the Federal Circuit 2026-02-27
• Judge panel: Not reached — no panel opinion on the merits
• Petition grounds: Challenged claims 1, 3–4, 6, 8–11, 13–24 under 35 U.S.C. § 103(a). Ground 1 — claims 1 and 3 obvious over Attard (US 9,406,177) + Frazer (US 9,494,926) (Frazer supplying the "fault-remediation table"); Ground 2 — claim 4 obvious over Attard + Hampiholi (US 2016/0267335); Ground 3 — claim 6 obvious over Attard + Hampiholi + McNew + Scofield; further grounds stacking DeRuyck, Gunderson, An, Kang, and Schunder.
• Institution decision: Denied. Per Tesla's cert petition, the Office expressly reasoned that "it is unlikely that a final written decision in this proceeding will issue before the district court trial occurs," and exercised discretion "based on a holistic assessment" notwithstanding Tesla's Sotera stipulation (and later broadened "Sotera Plus" stipulation, Dkt. 88, 2025-07-24).
• Final Written Decision: None issued. No claim of the '402 patent was canceled or upheld.
• Settlement / termination: None of record.
• Appeal: Yes, collaterally. Tesla petitioned the Federal Circuit for a writ of mandamus; denied 2026-02-27 in In re Tesla, Inc., No. 2026-116 (Fed. Cir.), on the authority of Mylan Laboratories Ltd. v. Janssen Pharmaceutica, N.V., 989 F.3d 1375 (Fed. Cir. 2021) (non-constitutional challenges to institution decisions not reviewable). Cert petition filed 2026-07-27, No. 26-136 (U.S.). Amicus briefs in support filed 2026-08-28 by Unified Patents, LLC and by a trade-association coalition (CCIA, HTIA, Alliance for Automotive Innovation, National Retail Federation).
• Defensive value: Zero estoppel and zero claim-level precedent. But the Attard + Hampiholi combination Tesla built for '402 claim 4 is a usable template for the competence-level comparison that is 9,566,986's core disclosure — see the inference flagged in the Strategic Summary.

IPR2025-00944 — Tesla, Inc. v. Granite Vehicle Ventures LLC (U.S. Patent 11,738,765)

• Type: Inter Partes Review
• Filed: 2025-05-02
• Status: Institution denied (same discretionary/time-to-trial basis); Federal Circuit mandamus denied 2026-02-27 under No. 2026-116
• Judge panel: Not reached
• Petition grounds: Challenged claims 1–2, 17, 6, 7, and others under § 103(a) across ten grounds — Ground 1 (claims 1–2, 17) over Attard + Frazer + McNew; Ground 2 (claim 7) over Hampiholi + Attard; Ground 3 (claims 3–4, 18–19) over Attard + Frazer + McNew + An + Kang + Schunder; Ground 4 (claims 5, 20) adding DeRuyck + Gunderson; Ground 5 (claim 6) over Attard + Frazer + Hampiholi; and so on through Ground 10 (claim 16).
• Institution decision: Denied. Patent Owner's opposition argued the grounds stacked "no less than four references, and as many as nine," and that no reference taught "active learning data" derived from "other SDVs."
• Final Written Decision: None issued. No claim canceled or upheld.
• Settlement / termination: None of record, but note Patent Owner filed a statutory disclaimer on 2025-08-08 (Ex. 2013 in the opposition record) in the '765 proceeding. This disclaimer concerns the '765 patent only — it does not affect 9,566,986. Do not conflate the two.
• Appeal: Covered by the consolidated mandamus/cert challenge described above.
• Defensive value: Same as above — no estoppel, no merits ruling. The useful takeaway is evidentiary: Patent Owner's own pleading concedes that every '765 claim requires "active learning data" derived from other SDVs, which is a narrowing construction argument worth reading before you draft art against 9,566,986.

IPR2025-01034 — Tesla, Inc. v. Granite Vehicle Ventures LLC (U.S. Patent 12,037,004)

• Type: Inter Partes Review
• Filed: 2025 (exact date not verified)
• Status: Institution denied (per the consolidated Federal Circuit mandamus order covering the family)
• Judge panel: Not reached
• Petition grounds: Eleven obviousness grounds under § 103(a) covering claims 1–9 and 29–39, combining Hampiholi, Attard, McNew, Frazer, DeRuyck, Gunderson, An, Kang, Schunder, Sako, Coelingh, Hada, Tsimhoni-647, Nagasaka, and Duncan.
• Institution decision: Denied — same discretionary basis.
• Final Written Decision: None issued.
• Settlement / termination: None of record.
• Appeal: Covered by No. 2026-116.
• Defensive value: The art list is the broadest in the family and is the best free prior-art roadmap available for the 2015-era self-driving-mode-switching space.

Two further Tesla petitions (numbers unverified)

Reporting indicates Tesla filed five IPR petitions against Granite's asserted patents. I have verified three numbers: IPR2025-00943, IPR2025-00944, and IPR2025-01034. I did not verify the remaining two numbers and will not guess at them. The Federal Circuit's caption in No. 2026-116 lists "IPR2025-00943, IPR2025-00944, IPR2025-…" truncated. All five were denied institution per the cert-stage briefing. If you need the complete set, pull the PTAB E2E docket for the family directly rather than relying on this list.

---

Strategic summary

Claim status on 9,566,986: everything is UNTESTED. Not a single claim of this patent is canceled, and not a single claim has been confirmed. Compare that to the rest of the family, where the only thing that has happened at the PTAB is that the Board declined to look — five times — and the Federal Circuit declined to make it look. So there is no "claims 1–5 are dead" argument and no "the patent is hardened" argument. There is simply a blank sheet, which cuts both ways: your opponent cannot point to a single adjudicated claim, and you cannot point to a single canceled one. The patent is active and, on Google Patents' face, runs to an anticipated expiration of 2035-09-25 — roughly nine years of runway, which is more than enough time to file, try, and appeal an IPR.

Estoppel landscape: essentially nothing binds anyone on this patent. Section 315(e)(2) estoppel attaches only after a final written decision, and no IPR on 9,566,986 has ever reached institution, let alone FWD. Tesla's denied petitions generated no estoppel — denial of institution does not trigger § 315(e). Non-parties (you, if you are not Tesla or a Tesla privy) are never estopped. If you are a Tesla privy, note that the estoppel question on this patent is vacuous because no proceeding existed; on the family patents, the non-institution likewise yields no estoppel, though the district court may still police Tesla's invalidity contentions under the case-management order in the transferred N.D. Cal. action. Practically: all prior-art grounds against 9,566,986 remain available to you. Your real obstacles are procedural, not estoppel-based — the § 315(b) one-year bar running from service of a complaint on you, and the current USPTO discretionary-denial posture that just defeated Tesla. That posture is the subject of a live Supreme Court cert petition; a favorable ruling in No. 26-136 would materially reopen the door for everyone.

Pattern signals. (1) One petitioner (Tesla) filed a coordinated multi-patent campaign against this family — five petitions, Sotera and "Sotera Plus" stipulations, then mandamus, then cert. That is an aggressive, well-funded defense, not a nuisance filing, and it tells you the family is genuinely vulnerable-looking to a sophisticated defendant. (2) Patent Owner is not pursuing PTAB appeals — there are none to pursue, because it won at the institution stage, and it apparently achieved that win partly by having the '765 claims narrowed by statutory disclaimer (2025-08-08). That is a meaningful tell: the '765 patent was disclaimed into a smaller footprint to survive. (3) Unified Patents is in the chain, but only as an amicus — it filed a cert-stage brief supporting Tesla on 2026-08-28 and maintains a public analytics page for US-9566986-B1. I found no Unified Patents IPR against 9,566,986. (4) The patent's own history is a monetization trail, not a practicing-entity trail: IBM → Daedalus Group LLC (recorded 2019-11-14) → Slingshot IOT LLC (2020-01-29) → Granite Vehicle Ventures LLC (2023-07-27), with Granite itself held by YEIP Holdings LLC and Albany Road Holdings LLC. (5) One inference you should treat as a lead, not a fact: Tesla's Ground 2 in IPR2025-00943 mapped Attard's overall confidence assessment ("Capital Φ") to a processor competence level and Hampiholi's driver-distraction "severity rank" to a human driver competence level. That is precisely the CPCL-vs-HDCL comparison that the 9,566,986 abstract describes. If the '986's independent claims read on that same comparison, an Attard + Hampiholi § 103 theory is a ready-made starting point. I have not read the '986 claim set in full and am not asserting that mapping is complete — verify it against the claims before you rely on it.

---

Recommended next steps

1. Do not represent to a court or an adversary that 9,566,986 has been invalidated at the PTAB. It has not. There is no FWD to cite, because no proceeding was instituted. If you want a PTAB outcome on this patent, you must create it.
2. File your own IPR, and watch the § 315(b) clock. If you have been served with a complaint asserting 9,566,986, you have one year from service. If you have not been served, you are not barred — and filing proactively is the only way to get a claim-level ruling. Budget for the discretionary-denial risk: build the petition to survive a Fintiv-style analysis, and consider whether a stipulation of the type Tesla used (and re-broadened) helps or hurts under current Office practice.
3. Screenshot and preserve the record now, while the cert petition is live. Track Tesla, Inc. v. Granite Vehicle Ventures LLC, No. 26-136 (U.S.), and In re Tesla, Inc., No. 2026-116 (Fed. Cir.). If the Supreme Court grants cert and reverses on the reviewability of extra-statutory institution denials, the calculus for filing against this entire family changes. Amicus support is substantial and bipartisan-industry (CCIA, HTIA, Alliance for Automotive Innovation, National Retail Federation, Unified Patents).
4. Pull the full PTAB E2E docket for the family yourself and independently confirm whether a sixth, unreported petition targeted 9,566,986. My search confirmed three petition numbers and failed to identify the other two of the five Tesla petitions; ODP ingest lag is a real possibility and I would not stake a filing decision on absence-of-evidence alone.
5. Mine Tesla's petition exhibits for art, but re-date them. Any reference you lift from the family IPRs (Attard, Frazer, Hampiholi, McNew, Scofield, DeRuyck, Gunderson, An, Kang, Schunder, Sako, Coelingh, Hada, Tsimhoni, Nagasaka, Duncan) must independently qualify as prior art as of 9,566,986's 2015-09-25 priority date. Note that Tesla expressly applied 2015-09-25 as the priority date while still relying on Hampiholi US 2016/0267335 — check its asserted effective filing date before you reuse that reference, because on its face a 2016 publication does not predate a 2015-09-25 filing.
6. Read the '765 statutory disclaimer and the '765 opposition record. Ex. 2013 (disclaimer, 2025-08-08) and the opposition's "active learning data" argument are the closest thing to a Patent Owner admission in this family, and they bear on how broadly Granite will be permitted to read the shared specification that 9,566,986 fathered.

---

Sources

• US 9,566,986 prosecution / bibliographic data and assignee chain — https://patents.google.com/patent/[US9566986B1](/patent/US9566986B1)/en
• Unified Patents public analytics page for US-9566986-B1 — https://portal.unifiedpatents.com/patents/patent/US-9566986-B1
• Granite's P.R. 3-1 infringement contentions (asserted claims identified; '986 not asserted) — https://ptacts.uspto.gov/ptacts/public-informations/petitions/[1557935](/patent/1557935)/download-documents?artifactId=h8TlvJEUMVHw7W-YpQ8NpQGcR4kLQQNWNZJLyKVgf_4lWy72jRq6SAQ
• IPR2025-00943 petition and grounds summary (Tesla v. Granite, '402 patent) — https://ai-lab.exparte.com/case/ptab/IPR2025-00943/doc/1006
• IPR2025-00944 grounds table and Patent Owner opposition ('765 patent) — https://ptacts.uspto.gov/ptacts/public-informations/petitions/[1557830](/patent/1557830)/download-documents?artifactId=pnhIxlYqra1KhrYZdCJUR4VKF_kwg_MBKCfFj1yxALW8iq8BCQqJPXE
• IPR2025-01034 case summary ('004 patent) — https://cases.arcticinvent.com/us/patent/us/petition-for-inter-partes-review-of-us-patent-no-12037004/ipr2025-01034/
• Tesla cert petition, No. 26-136 (U.S.), filed 2026-07-27 — https://business.cch.com/ipld/TeslaGraniteVehicleVentures26126PetitionWritCert20260727073126.pdf
• Unified Patents amicus brief, No. 26-136 — https://www.supremecourt.gov/DocketPDF/26/26-136/[422322](/patent/422322)/20260828074131666_Certificate%20of%20Word%20Count.pdf
• IPWatchdog, industry amici supporting cert (2026-08-31) — https://ipwatchdog.com/2026/08/31/computer-automotive-retail-industries-support-teslas-cert-petition-usptos-discretionary-ipr-denials/
• Transfer order, N.D. Cal. Case No. 26-cv-01457-TLT (TSH) — https://www.govinfo.gov/content/pkg/USCOURTS-cand-3_26-cv-01457/pdf/USCOURTS-cand-3_26-cv-01457-0.pdf
• Law360 docket page, Granite Vehicle Ventures LLC v. Tesla, Inc. — https://www.law360.com/cases/67536ec52ed14705899b498f

Confidence notes: The zero-proceeding finding for 9,566,986 rests on the ODP structured data plus my own searches, which I regard as high confidence but not absolute given possible ingest lag. The three family IPR numbers, their deny-institution outcome, and the Fed. Cir. denial (2026-02-27) are well corroborated across multiple sources. Specific filing dates for IPR2025-00944 (2025-05-02) and IPR2025-00943 (2025-05-05) come from the parties' filings and one secondary database. The exact filing date and two of the five petition numbers are unverified; I have not supplied numbers I could not confirm.

I'll start by pulling what's verifiable from the record and hunting for the recorded-assignment details (reel/frame, correspondent) that Google Patents' legal-events view doesn't show.

Strong leads on the chain. Now let me pin down reel/frame numbers, the correspondent of record, and the litigation timing.

Assignment & Ownership Analysis — US 9,566,986 B1

"Controlling driving modes of self-driving vehicles" · App. 14/865,393 · filed & issued from IBM Research

Sourcing note (read first): I could not open assignmentcenter.uspto.gov / assignment.uspto.gov directly from this environment, so I have not read the Assignment Center's own correspondent field. The reel/frame values below are taken from an on-record USPTO filing — a Statement Under 37 CFR 3.73(c) for this family, signed January 4, 2024 by /Jan D. MacKinnon/, which recites the chain by reel/frame — cross-checked against Google Patents legal events (https://patents.google.com/patent/[US9566986B1](/patent/US9566986B1)/assignments) and the RPX/Mondaq/IAM reporting on the IBM→Daedalus 2019 sale. Anything I could not verify is labeled as such rather than filled in.

---

Inventors

All five assigned to IBM on 2015-09-25 — the filing date itself — per the recorded inventor-to-IBM assignment (Google Patents legal events, 2015-09-25). Department-level affiliation is inferred from these inventors' well-known IBM Research roles, not independently verified here.

Unusual patterns — none detected, but a related note. I found no evidence that any inventor departed IBM within 12 months of filing; I also found no source that would let me verify their departure dates, so I am not asserting a fire-sale precursor. What is noteworthy is the assertion-driven continuation chain rather than inventor movement: continuations were filed 2016-11-02 ('225 → 10,029,701), 2018-04-18 ('874 → 10,717,446) during IBM ownership, then 2020-08-19 ('202 → 11,091,171), 2021-07-13 ('656 → 11,738,765), 2022-04-13 ('541 → 11,597,402), 2023-07-17 ('774 → 12,037,004) and 2024-06-14 ('841 → US 2025/0018963 A1) after the sale — i.e., the family was deliberately kept alive post-divestiture, with the newest continuation filed roughly six months before the December 2024 Tesla complaint.

Original assignee

International Business Machines Corporation (Armonk, NY) — named on the face of the issued patent; original assignee per the 2015-09-25 assignment and Google Patents.

• Product embodying the claims: No. The claims are directed to arbitrating control of a self-driving vehicle between an on-board SDV control processor and a human driver based on comparative competence levels. IBM never shipped a self-driving vehicle or a commercial SDV driving-mode controller; this is a research/portfolio asset from IBM's autonomous-vehicle and sensor-analytics work, not a product patent.
• Primary line of business: enterprise IT — hardware, software, cloud, and services; historically one of the largest US patent holders (~60,000 US patents at the time per IAM/ktMINE).
• Current status: Operating, solvent. No bankruptcy. The 2019 transfer was a negotiated divestiture — reported at $9 million for the largest tranche of 500+ US assets (plus European and Asian rights), with Houlihan Lokey advising IBM (IAM via Lexology, Nov 22 2019).

Assignment timeline

1. 2015-09-25 (executed) / recorded 2015-09-25 — Reel not captured in the sources I could reach

   • Conveyance: Assignment
   • Assignor: Michael S. Gordon; James R. Kozloski; Ashish Kundu; Peter K. Malkin; Clifford A. Pickover
   • Assignee: International Business Machines Corporation
   • Correspondent: not verifiable from my sources
   • Context: routine employee-invention assignment, executed on the filing date.

2. 2019-09 (executed per RPX/IAM) / recorded 2019-11-14 — Reel 051032/0784

   • Conveyance: Assignment — this recording is the complete IBM–Daedalus transaction agreement, not a short-form assignment. IBM and Daedalus later petitioned the USPTO to expunge it as erroneously recorded on November 14, 2019, at Reel/Frame 051032/0784, and to replace it with the conforming assignment (see the expungement petition in App. 16/172,085). The same reel/frame covers this family's tranche.
   • Assignor: International Business Machines Corporation
   • Assignee: Daedalus Group LLC (Delaware; principal Ed Gomez — ex-Altitude Capital, Walker Innovation, General Patent Corporation; address associated with Gomez in Bronxville, NY)
   • Correspondent: not verifiable; the expungement petition itself was filed on behalf of Daedalus as applicant
   • Context: divestiture / monetization vehicle — 500+ US assets plus foreign counterparts, $9M for the largest tranche.

3. 2020-01-27 (recorded) — Reel 051710/0445

   • Conveyance: Assignment (the conforming replacement assignment for the same IBM→Daedalus transaction, intended to supersede the accidentally recorded agreement)
   • Assignor: International Business Machines Corporation
   • Assignee: Daedalus Group, LLC
   • Correspondent: not verifiable
   • Context: same transaction, re-recorded — this is why Google Patents shows two IBM→Daedalus events (2019-11-14 and 2020-01-27). Both reels 051032/0784 and 051710/0445 appear as consecutive entries in this family's §3.73(c) chain-of-title statement.

4. 2020-01-29 (recorded) — Reel 051733/0463

   • Conveyance: Assignment
   • Assignor: Daedalus Group, LLC
   • Assignee: Slingshot IOT LLC (Baltimore, MD)
   • Correspondent: not verifiable
   • Context: transfer-to-asserter vehicle — RPX reports Daedalus handed ~75 assets from the IBM tranche to Slingshot IOT, a Maryland entity managed by York Eggleston and Keith Machen as "part of a growing web of assertion entities" (RPX/Mondaq, July 22 2020). Reel 051733 lands three days after 051710, consistent with the same-week recording burst.

5. 2023-07-27 (per Google Patents legal events) — Reel/frame NOT VERIFIED

   • Conveyance: Assignment
   • Assignor: Slingshot IOT LLC
   • Assignee: Granite Vehicle Ventures LLC (Carrollton, TX — address shown in the family's ADS of record, e.g. the '004 continuation filing)
   • Correspondent: Ian D. MacKinnon signed this family's §3.73(c) chain-of-title statement on January 4, 2024, and is the filer of record on the 2021 continuation papers (Customer No. 164). I could not confirm whether MacKinnon is also the named correspondent on the assignment recordings themselves — flagging as unresolved rather than asserting recurrence.
   • Context: transfer-to-asserter. Granite is the current assignee per Google Patents and per the USPTO-facing ADS; Unified Patents lists its "Parent Company" as Granite Vehicle Ventures LLC and describes it as an NPE (Unified portal).

Records exist, so the "no assignments recorded" branch does not apply.

Timeline diagram

timeline
    title Ownership of US 9566986
    2015 : Filed by IBM inventors
         : Assigned to IBM
    2017 : Patent issued 14 Feb
    2019 : IBM sells to Daedalus Group
    2020 : Daedalus assigns to Slingshot IOT
    2023 : Slingshot IOT to Granite Ventures
    2024 : Granite sues Tesla
    2025 : Tesla files five IPR petitions

NPE / troll-pattern signals

1. Shell-entity transfer — PRESENT. The patent leaves an operating company (IBM) and passes into a three-stage chain of non-practicing entities with no products in commerce: Daedalus Group LLC (Reel 051032/0784; 051710/0445), Slingshot IOT LLC (Reel 051733/0463), and Granite Vehicle Ventures LLC (2023-07-27, reel unverified). Naming tells are present ("Group LLC," "IOT LLC," "Ventures LLC"), but the concrete evidence is stronger than the naming: RPX describes Daedalus principal Ed Gomez's IP-monetization-only background and the Eggleston/Machen web of assertion entities, and Unified Patents labels Granite an NPE whose sole business is asserting former-IBM SDV patents. Granite's USPTO address of record is Carrollton, TX.

2. Known asserter in the chain — PRESENT. Not Acacia, Marathon, IV, Wi-LAN etc. — but the chain is populated by entities with documented assertion campaigns: Daedalus Blue (affiliate of the Reel 051032/0784 assignee) sued SZ DJI in W.D. Tex. and Oracle (6:20-cv-00428) within months of the transfer (Khurana & Khurana); Slingshot IOT's sibling entities (Slingshot Printing, Quartz Auto Technologies, Terrace Licensing, Stone Interactive Ventures, Jawbone Innovations) are active plaintiffs; and Granite Vehicle Ventures sued Tesla in December 2024, drawing five IPR petitions (e.g. IPR2025-00943 on 11,597,402 and IPR2025-01034 on 12,037,004 — both in this same family), a Unified Patents PATROLL $2,000 prior-art bounty on 12,037,004, and a Tesla cert petition, Tesla, Inc. v. Granite Vehicle Ventures LLC, No. 26-136.

3. Repeat correspondent across the chain — UNCLEAR / not established. The one recurring attorney I can document on this family is Ian D. MacKinnon (Customer No. 164): filer on the 2021 continuation papers for App. 17/374,656 and signer of the January 4, 2024 §3.73(c) chain-of-title statement covering Reels 051032/0784, 051710/0445 and 051733/0463. That is a prosecution/filing recurrence across the Slingshot→Granite era, which is suggestive but not the same as the assignment-record correspondent field. I did not verify the Assignment Center correspondent for any of the four recordings, so I decline to call this signal present. (Customer No. 164 is a shared attorney customer number rather than an individual filer identity — worth an Assignment Center pull to resolve.)

4. Cascading transfers — PRESENT (partial). Two rapid hops: IBM→Daedalus recorded 2019-11-14 at 051032/0784 (conforming re-recording 2020-01-27 at 051710/0445), then Daedalus→Slingshot IOT recorded 2020-01-29 at 051733/0463 — three reels inside ~10 weeks, on a single corporate address (Bronxville, NY for Daedalus/Ginegar). The third hop, Slingshot IOT→Granite, did not occur until 2023-07-27, a ~3.5-year hold; so the cascade is two-stage, not three-stage-under-24-months.

5. Pre-litigation transfer — NOT PRESENT as to the 6-month window. The Granite acquisition (2023-07-27) precedes the December 2024 Tesla complaint by roughly 17 months — outside the 6-month tell. Note however that the family's newest continuation was filed 2024-06-14, about six months before the complaint, consistent with a deliberate pre-suit portfolio build. The assignment itself, as to this window, does not qualify.

6. Bankruptcy fire-sale — NOT PRESENT. IBM is solvent and operating; the transfer was a negotiated, advised divestiture at a stated $9M for the largest tranche, with IBM retaining ~60,000 US patents. This is not a Kodak/Nortel/Polaroid scenario.

7. Privateering — UNCLEAR (partial evidence). The reported deal documents identify which companies were licensed and unlicensed to the transferred assets, and the assets were then asserted against Oracle, DJI, Uber, Lyft and Tesla. IBM also joined the LOT Network in January 2020, described by LOT as following the Daedalus sale by about two weeks (LOT Network) — immunizing IBM and other LOT members while asserting against non-members. That pattern is privateering-adjacent, but I found no evidence IBM directed or benefited from the campaigns, so I do not score this as present.

8. Defensive aggregator — NOT PRESENT. The chain terminates at Granite Vehicle Ventures LLC, an asserting NPE litigating against Tesla, not at RPX, AST, LOT, Unified or OIN. The inverse signal applies: this patent is live and weaponized, not neutralized.

Verdict

NPE — high confidence.

Three independent strong signals converge on the same conclusion. First, shell-entity transfer: the asset exited IBM and moved through an unbroken chain of non-practicing entities — IBM → Daedalus Group LLC (Reel 051032/0784 recorded 2019-11-14, conforming assignment at Reel 051710/0445 recorded 2020-01-27) → Slingshot IOT LLC (Reel 051733/0463 recorded 2020-01-29, the Eggleston/Machen assertion web) → Granite Vehicle Ventures LLC (2023-07-27). Second, known asserter in the chain: Daedalus Blue, an affiliate of the same tranche purchaser, sued DJI and Oracle, and Granite Vehicle Ventures — the current assignee of record — sued Tesla in December 2024 on this family's continuations (IPR2025-00943 on 11,597,402; IPR2025-01034 on 12,037,004), with Unified Patents expressly designating Granite an NPE. Third, cascading transfers: two conveyances across three reels inside roughly ten weeks (051032/0784 → 051710/0445 → 051733/0463). Bankruptcy (not present) and defensive-aggregation (not present) do not soften the call; the only signal I could not confirm is repeat assignment-record correspondent, because the Assignment Center's correspondent field was not reachable from this environment.

Verify at: https://assignmentcenter.uspto.gov/ (search patent 9,566,986) — confirm the correspondent of record for Reel 051032/0784, 051710/0445 and 051733/0463, and obtain the reel/frame for the 2023-07-27 Slingshot IOT → Granite Vehicle Ventures assignment, which is the one gap in this chain.

I'll search for the specific prior-art citations listed on the face of US 9,566,986.

Let me dig into the specific front-page references and the "Patent Art" list.

Let me pull the granted PDF front page and the citation lists.

Prior Art for US 9,566,986 B1 — Analyst Report

Date of analysis: 2026-10-01 · Patent: US 9,566,986 B1 · App. 14/865,393 · Effective filing date: 2015-09-25

---

Important sourcing caveat — read this first

I could not retrieve the itemized "References Cited" table as it appears on the front page of the US 9,566,986 grant, and I am telling you that rather than reconstructing it from memory or from a sibling patent. Here is precisely why:

1. The authoritative full text supplied for this analysis (the Google Patents fetch) is truncated inside the Detailed Description, before the claims and before any citation table. It contains no front-page "References Cited" list.
2. Google Patents' rendered page for US9566986B1 does expose citation data programmatically, but my fetches returned the description/citation-count fragments, not the itemized reference rows.
3. What I did retrieve reliably is a count summary, from Unified Patents' page for US-9566986-B1: Patent Art (76) · Non-Patent Literature (23) · Referenced By (90). That tells you the '986 patent carries 76 patent references and 23 NPL items on its face, but not what they are. (https://portal.unifiedpatents.com/patents/patent/US-9566986-B1)

So the sections below are organized in descending confidence. §1 is what I can state about the citation record itself. §2 is the most relevant candidate art — art I surfaced from the '986 family's examiner search string, from sibling front pages, and from Tesla's family IPRs — with an explicit confidence flag on each. §3 is the §102 framework and element-by-element analysis. §4 tells you how to close the gap definitively in one step. I have not guessed at any reference I could not at least locate a number for.

Two date notes that matter for every §102 line below: (a) the effective filing date is 2015-09-25 per Google Patents, but Unified Patents records a 2015-09-24 priority date — the one-day discrepancy flagged in the earlier bibliography section, which I do not resolve here; and (b) because the application was filed after 2013-03-16, AIA 35 U.S.C. §102(a)(1)/(a)(2) governs, not pre-AIA §102(a)/(b). "Anticipation" therefore means a single reference that was (1) publicly available before 2015-09-25, or (2) a US patent/published application effectively filed before 2015-09-25 (issue date is irrelevant under (a)(2)).

---

1. What is verified about the citation record

What I could NOT obtain: the itemized 76-reference list, the 23 NPL items, and the specific patentee/examiner attribution of each. I will not manufacture these.

One point worth flagging for a §102 challenge: under AIA §102(a)(2), a reference must "name another inventor." The '986 specification fathered a large IBM sibling family with the same five inventors (Gordon, Kozloski, Kundu, Malkin, Pickover). Sibling disclosures by that identical inventive entity are therefore not §102(a)(2) prior art against '986 — and to the extent any are commonly owned by IBM at the time, §102(b)(2)(C) excludes them independently. That removes an entire class of superficially attractive candidates.

---

2. Most relevant candidate prior art

2A. Candidates surfaced from the '986 family's own examiner search string (highest relevance to this disclosure)

A USPTO examiner search string appears in a PTAB exhibit for this family (L17, run 2022/07/21), enumerating the art the examiner treated as the relevant field for the "controlling driving modes of self-driving vehicles" family. It is the single best proxy I found for what the '986 examiner considered. The list is:

US 8,874,301 · US 9,135,803 · US 9,238,467 · US 9,566,986 · US 9,663,118 · US 9,823,657 · US 2012/0046817 · US 2014/0303827 · US 2015/0066284 · US 2015/0088357 · US 2015/0149021 · US 2015/0175070 · US 2015/0314780 · US 2016/0187879 · US 2016/0202700 · US 2016/0209840 · US 2017/0038773 · US 2017/0115661 · US 2017/0227959 · US 2017/0247041 · US 2017/0364070 · US 2018/0039268 · US 2018/0043904 · US 2018/0050704 · US 2018/0118219

(Source: PTAB petition exhibit, ptacts.uspto.gov/.../1557829/..., lines L13–L17.)

Caveats — please read. This is an examiner's search query, not a confirmed front-page citation list. It mixes (i) pre-2015-09-25 art that can anticipate, (ii) post-filing publications that can only qualify under §102(a)(2) if effectively filed earlier, and (iii) the '986 patent itself (used as an "adjacent" hit). Two entries are notable and appear to be the same references Tesla later used against the continuations: US 2015/0175070 A1 (Attard et al.) and US 2015/0149021 A1 (Duncan et al.).

Selected entries, with what I can responsibly say:

2B. Boilerplate references shared across the IBM SDV family front pages

The sibling IBM grants in this same drafting family ("Controlling driving modes of self-driving vehicles," Law Office of Jim Boice) list a common block of older art. Because the same firm drafted '986 and its siblings from a shared specification, these are probable — but unconfirmed — members of the '986 front page:

• US 4,665,395 A — Van Ness, 5/1987 (confirmed present on US 9,785,145 and US 9,834,224 front pages)
• US 4,908,988 A — Yamamura et al., 3/1990 (US 9,785,145) / US 4,905,798 A — Yamamura et al., 3/1990 (US 9,834,224)
• US 5,975,791 A — McCulloch, 11/1999
• US 6,064,970 A — McMillian et al., 5/2000
• US 6,201,318 B1 — Guillory, 3/2001
• US 6,326,903 B1 — Gross et al., 12/2001
• US 6,393,363 B1 — Burns, 5/2002
• US 6,502,035 B2 — Levine, 12/2002
• US 6,587,043 B1 — Kramer, 7/2003
• US 6,622,082 B1 — Schmidt et al., 9/2003
• CN 1135063 (11/1996); CN 2349068 Y (11/1999)
• NPL: P. Mell et al., "The NIST Definition of Cloud Computing," NIST, Sep. 2011 (cited for the cloud-computing boilerplate in the specification) and, on the siblings, a "U.S. Appl. No. 14/855,731 Non-Final Office Action" cross-reference.

§102 assessment of this block. These are largely pre-1990-to-2003 vehicle-control and monitoring references that could at most touch ancillary elements (sensor detection of vehicle faults, driver monitoring). None of them, individually, discloses the core '986 combination — an on-board autonomous control processor's competence level computed for a specific operational anomaly and compared against a human driver's competence level for the same anomaly, with control assigned to the higher. I would not plead any of these as an anticipating reference; they are §103 background at best. Confidence that they are on the '986 front page: moderate; confidence that they anticipate: low.

2C. Art from the family IPRs (roadmap, not adjudication)

The earlier sections of this analysis document that Tesla's IPR petitions against the continuations (US 11,597,402; 11,738,765; US 12,037,004) stacked: Attard, Frazer, Hampiholi, McNew, Scofield, DeRuyck, Gunderson, An, Kang, Schunder, Sako, Coelingh, Hada, Tsimhoni-647, Nagasaka, Duncan. All five petitions were denied institution — so none of this art has been adjudicated, and none of it creates estoppel. It is nonetheless the best free art roadmap available for this family. Two cautions before you reuse it:

• Re-date every reference to 2015-09-25. In particular, Tesla relied on Hampiholi, US 2016/0267335, which on its face is a 2016 publication; it can only reach '986 under §102(a)(2), which requires an effective filing date before 2015-09-25. Verify that before use — the earlier section flagged this exact issue.
• Frazer (US 9,494,926) was used by Tesla to supply the "fault-remediation table" element. If the '986 independent claims recite that table, Frazer is a directly relevant secondary reference.

§102 note: none of these was applied to '986 itself. They are §103 candidates against the '986 claims; for §102 anticipation you need a single reference teaching all elements.

2D. '986's own family (not prior art to itself)

For completeness and to prevent conflation: '986's pre-grant publication is US 2017/0088145 A1 (published 2017-03-30); its continuations are US 10,029,701; 10,717,446; 11,091,171; 11,738,765; 11,597,402; 12,037,004; and pending US 2025/0018963 A1. A patent's own family is not §102 art against it, and the same-inventor/common-ownership exclusions in §102(b)(2)(A)/(C) independently disqualify the IBM siblings.

---

3. §102 framework applied to the '986 independent claims

Caveat carried forward from the earlier section: I could not obtain the verbatim granted claim set of '986, so the element list below is drawn from the patent's own Summary and abstract (itself authoritative for scope) rather than from parsed claim text. Treat the mapping as element-level against independent claim 1 (method) and its system and computer-program-product counterparts, which I assume exist based on the Summary's "method, system, and/or computer program product" framing. Do not cite specific dependent-claim numbers from this report.

For a single reference to anticipate under AIA §102(a)(1) or (a)(2), it must disclose all of:

1. Receiving sensor readings describing a current operational anomaly of an SDV traveling on a roadway;
2. The SDV being operable in autonomous mode by an on-board SDV control processor;
3. A driving mode module that selectively controls autonomous vs. manual mode (manual = human driver controls);
4. Determining a control processor competence level of the on-board processor for that anomaly;
5. Receiving a driver profile giving a human driver competence level for that anomaly;
6. Comparing the two competence levels; and
7. Selectively assigning control to processor or human based on which level is relatively higher.

Anticipation screen:

• Attard (US 2015/0175070 A1) is the closest thing I found to a single-reference §102 candidate. It plausibly supplies elements 2, 3, 4 and a form of 7 (processor-confidence-driven handover). Element 5 is the likely gap: I could not confirm Attard compares a human driver competence level derived from a driver profile against the processor level for the specific anomaly. If Attard lacks element 5, Attard alone does not anticipate, and the correct posture is §103 over Attard + a driver-profile/secondary reference (e.g., Hampiholi, or one of the 2015 driver-state references above).
• The pre-2015-09-25 publications in the examiner search string (US 2015/0066284; 2015/0088357; 2015/0149021; 2015/0175070; 2014/0303827; 2012/0046817) are each §102(a)(1)-eligible on date. I have not verified their content, so I do not assert that any anticipates.
• Everything issued 2016 or later (US 9,238,467; 9,663,118; 9,823,657; the 2016–2018 publications) is eligible only under §102(a)(2), and only if you confirm the reference's effective filing date precedes 2015-09-25. Issue date is legally irrelevant for that purpose.

The realistic §102 answer: on the evidence I can actually see, the '986 independent claims are more likely vulnerable to §103 obviousness than to single-reference §102 anticipation — the "compare processor competence to human competence and pick the higher" arbitration is the piece the 2015-era art tends not to show in a single document. That is consistent with the family history the earlier sections document: Tesla pleaded §103 only, never §102, against the continuations.

---

4. How to close the gap — the one authoritative step

The front-page citation list is a matter of public record and retrievable in one move. Do not rely on this report's §2B block or the §2A search string as the citation list:

1. Open the granted-PDF images for US 9,566,986 via USPTO Patent Public Search / Patent Center (https://ppubs.uspto.gov or https://patentcenter.uspto.gov) and read the "(56) References Cited" column on the face of the patent. That is the definitive itemization of all 76 patent references and 23 NPL items, with examiner/patentee attribution.
2. Or use Google Patents' structured citation tabs for US9566986B1 — "Patent Citations," "Non-Patent Citations," and "Cited By" — which will render the rows that my fetches could only summarize as counts.
3. Then re-screen each reference against the 2015-09-25 effective date, splitting them into §102(a)(1)-eligible (published before) and §102(a)(2)-only (US patent/pub effectively filed before), and drop any IBM sibling with the same five inventors.
4. Cross-check the PTAB E2E docket for the family in case the two unverified Tesla petition numbers (only three of five confirmed) targeted '986 — the earlier section flags ODP ingest lag as a real risk.

---

Sources

• Count summary — Patent Art (76) / Non-Patent Literature (23) / Referenced By (90) — https://portal.unifiedpatents.com/patents/patent/US-9566986-B1
• Examiner search string (L13–L17, 2022-07-21) from a family PTAB exhibit — https://ptacts.uspto.gov/ptacts/public-informations/petitions/[1557829](/patent/1557829)/download-documents?artifactId=BbDKGzqlRQcneJL5kJQYYzqJyL9LH_68BNrIu1LpYCcqtRGOOh8o7hk
• Sibling front-page reference blocks (shared drafting family) — US 9,785,145 B2: https://patentimages.storage.googleapis.com/e0/e6/24/9862d271c1f5a3/US9785145.pdf · US 9,834,224 B2: https://patentimages.storage.googleapis.com/e7/ea/45/65f3423ffcc47c/US9834224.pdf
• '986 family continuation chain (app. 14/865,393 → PAT 9566986) — https://ptacts.uspto.gov/ptacts/public-informations/petitions/[1557935](/patent/1557935)/download-documents?artifactId=Ftf-xGvmEVcpm11-eB3zW9Ck6oB4UfvvLKwh-lFs6x9NL4XRMbwk2IE
• Family IPR art (Attard/Frazer/Hampiholi et al.) — IPR2025-00943: https://ai-lab.exparte.com/case/ptab/IPR2025-00943/doc/1006 · IPR2025-00944: https://ptacts.uspto.gov/ptacts/public-informations/petitions/[1557830](/patent/1557830)/... · IPR2025-01034: https://cases.arcticinvent.com/us/patent/us/petition-for-inter-partes-review-of-us-patent-no-12037004/ipr2025-01034/
• Authoritative full text (truncated before claims/citations) — https://patents.google.com/patent/US9566986B1/en

Confidence summary: High — that I could not retrieve the verbatim front-page list, and the 76/23/90 counts. Moderate — that the §2A search-string entries reflect the examiner's actual field of search; that Attard US 2015/0175070 A1 is the closest single-reference §102 candidate. Low / unverified — the specific titles, dates, and content of most individual references, and any assertion that a particular reference anticipates a particular claim. I have not fabricated a single citation in this report.

I'll ground the prior-art references before drafting the obviousness analysis.

Let me pin down dates and remaining references, and make one more attempt at the '986 claim set.

Obviousness Analysis — US 9,566,986 B1 under 35 U.S.C. § 103

Note on what this section builds on: I do not repeat the bibliographic record, assignment chain, litigation posture, or PTAB history from the earlier sections. Those are cross-referenced where they bear on the § 103 analysis (chiefly the art set Tesla assembled in the family IPRs and the priority-date discrepancy flagged in § 1 of the summary).

---

1. Threshold caveat — read before relying on any mapping below (fatal-if-ignored)

I still do not have the verbatim granted claim set of US 9,566,986. The authoritative full text I was given is truncated in the detailed description before the claims, and my searches returned claim language only for the continuation patents (US 11,597,402, US 11,738,765, US 12,037,004). Accordingly:

• Everything below is keyed to the '986 disclosure's own Summary and abstract, which recite the same seven-step architecture that the family's continuation claims chart out. That is a characterization, not a quotation, of the '986 claims.
• If the '986's independent claims differ materially from the family claim language — e.g., if they lack the "comparison" step, or if they claim the fault-remediation table in the independent claim rather than a dependent — the mapping below must be re-run against the actual claims. The scope of the art will not change; the element the art must reach will.
• I flag explicitly that all family IPR petitions were denied institution, so there is no Board claim construction, no FWD, and no estoppel anywhere in this family. Everything here is first-principles analysis, not adjudicated record.

One open item from the earlier sections is now resolved (and I am flagging it because I earlier said I could not resolve it): the PTAB section noted that Hampiholi (US 2016/0267335), a 2016 publication, could not on its face predate the '986's 2015-09-25 filing. The record shows Hampiholi was filed 13 March 2015 as Application 14/657,070. Under AIA § 102(a)(2) a U.S. application publication is prior art if it was "effectively filed" before the claimed invention's effective filing date, regardless of its publication date. Filed 2015-03-13 < 2015-09-25 → Hampiholi qualifies as § 102(a)(2) art. Not commonly owned (Harman International vs. IBM/Granite), so § 102(b)(2)(C) does not disqualify it. The date problem I previously flagged is not a problem.

Contradiction flagged, not resolved: Google Patents gives priority/filing/expiration as 2015-09-25 / 2017-02-14 / 2035-09-25; Unified Patents gives 2015-09-24 / 2017-02-13 / 2035-09-24. For § 103 purposes the one-day delta is immaterial — Attard published broadly on 2015-06-25 and again 2015-09-23, and Hampiholi was filed 2015-03-13. I use 2015-09-25 as the effective filing date throughout, consistent with the family's own infringement contentions and the Office's treatment, and flag that a defendant who needs the earlier date must plead it.

---

2. Governing framework

• The '986 was filed 2015-09-25, after 16 March 2013 → AIA § 102/§ 103 apply. Prior art = anything publicly available before 2015-09-25 (§ 102(a)(1)) or any patent/application publication effectively filed before that date (§ 102(a)(2)).
• Obviousness asks whether the subject matter as a whole would have been obvious to a POSITA. Graham v. John Deere, 383 U.S. 1 (1966) (scope/content, differences, PHOSITA level, secondary considerations); KSR Int'l v. Teleflex, 550 U.S. 398 (2007) (predictable combinations; "obvious to try"; design incentives and market forces as motivation).
• The "reason to combine" can come from the references themselves, from the knowledge of a POSITA, or from the nature of the problem (here, the well-documented Level-3 handover problem). KSR; In re Kao, 639 F.3d 1057 (Fed. Cir. 2011).
• All references below are analogous art — same field of endeavor (vehicle control/ADAS), and reasonably pertinent to the same problem (who should drive when the automation's capability degrades).

POSITA definition I apply (matching what the family petitions used): a bachelor's degree in mechanical, electrical, computer engineering or a related field, plus ~2–5 years of experience in vehicle control systems, autonomous/ADAS systems, or the equivalent. This is the level at which the combination below is evaluated.

---

3. The elements that must be reached (reconstructed from the '986 Summary/abstract)

---

4. Prior-art qualification

---

5. GROUND 1 (primary): Attard + Hampiholi

This is the combination that maps most directly onto E1–E7, and it is the same pairing Tesla built for family claim 4 in IPR2025-00943 (mapping Attard's Φ to a processor competence level and Hampiholi's R to a human driver competence level).

Motivations a POSITA would have had (the KSR rationales, in the order I would plead them):

1. Express lead in Attard. Attard's parameter 117 is already "associated with a particular vehicle 101 operator," i.e., Attard already individualizes its autonomy decision to the driver. Substituting/adding Hampiholi's driver-derived scalar R into that parameterization is a substitution of one known element for another to obtain the predictable result of a better-tailored handover decision. KSR.
2. The references address complementary halves of one problem. Attard monitors whether the vehicle should be driving; Hampiholi monitors whether the human is fit to drive. Both target the same safety objective, and the Level-3 handover problem was a recognized, express industry concern by 2014–2015 (SAE J3016, Jan. 2014; Meyer 2014, at 50/56/204). Combining vehicle-side and driver-side assessments is the obvious design response to a recognized problem, not a new insight.
3. Structural and hardware compatibility. Both systems run on a vehicle computer with processor, memory, cameras, and vehicle-network data. Attard's vehicle is natively manually drivable (2:55-3:4), so no new hardware (steering wheel, pedals) is needed. The modification is straightforward programming — collapse two scalars and compare them. Predictable result, high REOS.
4. Improvement of the same device in the same way. Hampiholi's R was computed and ranged to select an action; Attard's Φ was computed and thresholded to select an action. Applying the known Hampiholi ranging technique to Attard's decision improves accuracy and appropriateness of Attard's corrective action in exactly the way it improved Hampiholi's. KSR rationales (C) and (D).
5. Design incentive / market forces. Federal policy activity around automated vehicles (NHTSA policy statements; Harding et al. DOT HS 812 014, Aug. 2014) and commercial pressure from the DriveMe-type programs created a strong incentive to solve the driver-handover problem — the classic "design incentive" motivation.
6. No teaching away. Neither reference disparages driver monitoring or vehicle-confidence assessment; Hampiholi's [0002]/[0017] expressly frame the goal as reducing crash risk, which is furthered by also detecting vehicle-side faults.
7. Obvious to try. Even on a stricter view, a POSITA faced with the handover problem had a finite set of predictable options (trust the car, trust the human, or measure both and pick), and measuring both was a known, finite, and predictable route.

Anticipated Patent Owner rebuttals, and my honest assessment:

• "Attard's Φ is a confidence in sensor/perceptual data, not a competence level of the control processor." This is the strongest counter. But Attard's Φ is described as bearing on whether the vehicle "can and/or should be operated autonomously," which is functionally a capability statement about the autonomous controller. Claim-construction risk: moderate. The family's own petition success at the institution stage was not on the merits, so nothing is adjudicated.
• "Hampiholi's R is a distraction severity rank, not a competence level." Weaker counter — R is derived from driver state and is expressly a graded measure of the driver's fitness to drive in context. But if the claims require an explicit "competence" metric with a driver profile, Hampiholi alone is thin; add the driver-profile art in Ground 5.
• "Neither reference compares the two levels to each other; Attard compares Φ to a fixed φmin, Hampiholi compares R to fixed ranges." This is the genuine vulnerability. See § 12 below. Mitigate with An (risk-index that combines driver and vehicle inputs) or with DeRuyck/Gunderson/Coelingh, or argue that "relatively higher" reads on normalized-score comparison.

---

6. GROUND 2: Attard + Frazer (fault-remediation table — D1)

Frazer supplies the element the '986 discloses as a mapping of "fault condition → manifested condition → required driving mode," updated from a central server. Frazer's event table 112 contains rows 140c (faults) with a condition column 142c and an instruction/action column 144c, and Frazer expressly claims low latency (23:64-67) — an express reason to prefer a table over ad hoc conditional code.

• Motivation: routine data structuring for retrieval speed and reduced memory vs. an unorganized list; Frazer's own latency teaching is a reference-based reason to combine.
• REOS: high — this is a data-storage-arrangement change to Attard's existing memory, implemented by programming. No bodily incorporation, no new hardware.
• Use it for: D1, and as the organizing structure vehicle for D2 (a "weighted voting parameter" can be one of the table's columns).

---

7. GROUND 3: + McNew (and Scofield) — actual handover, alerts, driver-state sensing (E3, E7, D3)

Attard's message 116 recommends manual control; McNew supplies the mechanism for actually performing the state transition and for sensing driver states (HANDS_ON / HANDS_FREE / DRIVER_IS_STEERING via grip sensor), plus the 1.5-second driver response-time figure (McNew 4:47-61) that quantifies why the handover decision matters.

• Motivation: Attard's repeated express disclosures recommending transfer to manual (2:55-3:4) are a specific suggestion; McNew supplies the known technique for executing it and adds driver-state confirmation. Result: the SDV both alerts and transitions. Scofield supplies surveillance-of-the-driver-for-handover.
• Use it for: E7, D3, and the "driver profile" flavor of E5.

---

8. GROUND 4: + An (and/or Kang) — weighted voting and weighted-summed sensor readings (D2)

An (US 9,063,543) teaches cooperative autonomous driving "between a vehicle and a driver" with a risk index built by weighting inputs. Attard itself teaches weighting a plurality of determinations, and An supplies the explicit weighting formulation; Kang supplies weights varying by weather condition.

• Motivation: weighting inputs improves accuracy and flexibility of the same computation (Attard's Φ); a POSITA would weight road-condition data and precipitation data by their relative impact on autonomous capability so that the more consequential condition dominates the mode decision. Weighting equation inputs was notoriously well known — a "known technique used to improve a similar device in the same way."
• Use it for: D2, and to support the summing and predefined level limitations of the weighted-sum embodiment.

---

9. GROUND 5: driver-profile and processor-cohort elements (E5, D7)

This is where the theory is thinnest and where I would expect a Patent Owner to concentrate. The '986's cohort/active-learning disclosure (drivers or processors sharing traits with a cohort whose accident record is known; prohibiting manual mode when the cohort's accident rate exceeds the processor's) is the most distinctive part of the specification — and, tellingly, it is the limitation Patent Owner itself argued in the '765 opposition ("active learning data" from "other SDVs") and the reason the '765 claims were narrowed by statutory disclaimer on 2025-08-08. That is a meaningful signal that this feature is the family's narrow point.

Available art: Meyer (fleet/handover surveillance, driver-out-of-the-loop at 50/56/204); Harding et al. / NHTSA (fleet-level V2V data); Mosquet et al., BCG (April 2015) (fleet and insurance data on autonomous deployment); Attard's own V2V communication 112 and network 120/remote server 125 (fleet data sharing, which is the mechanism for cohort learning); Tsimhoni-323; statistical/machine-learning aggregation generally.

Honest assessment: Attard's V2V/network/server architecture plus active learning is a recognized, well-known technique gives a defensible § 103 argument, but it is a weaker ground than Grounds 1–4 because no verified reference expressly teaches cohort-matched accident statistics driving a mode prohibition. If the '986's independent claims contain the cohort/active-learning limitation, this is where you either (a) find a better reference, or (b) rely on the obvious-to-try rationale that collating fleet outcomes to tune a threshold was a known, finite practice. Do not overstate this ground.

---

10. GROUND 6: environmental-report / sensor-conflict; anomaly enumeration (D5, D6)

• D6 (weather service trusted over on-board sensors): Attard expressly contemplates external networks and servers (120/125) and V2V data 112, and teaches that environmental conditions ("wet or icy roads," 7:14-17) affect confidence. Dannheim et al. (2014) teaches camera/LIDAR weather detection; Harding et al. (NHTSA, 2014) covers V2V data availability; Tsimhoni-323/Sato cover infrastructure-sourced conditions. Motivation: an on-board sensor samples only the surface beneath the vehicle at an instant, whereas a regional report covers the roadway ahead — a classic reason to prefer the broader data source. Strong ground.
• D5 (enumerated vehicle anomalies): Attard's perceptual-layer/actuation-layer confidence model covers sensor and actuation degradation, including braking; Muldoon et al. (IEEE 2002) teaches sensor-fusion vehicle fault diagnostics; Caveney and Stenneth cover vehicle/road condition data; standard OBD-II diagnostics were ubiquitous and expressly referenced in the Attard-family art. Snow tires / tread / washer fluid / wiper edge / headlamp are all matters of routine vehicle condition sensing — enumerated limitations that a POSITA would implement with known sensors. Strong ground.

---

11. GROUND 7: roadway width and geometry (D8)

The '986 makes the assignment depend on roadway width and on roadway geometry (the cloverleaf example). Attard expressly teaches detecting road attributes with its data collectors — "curves, potholes, dips, bumps, changes in grade, lane boundaries" — and using them in confidence assessment, which subsumes roadway geometry; lane width is a lane-boundary attribute. Cullinane (US 8,825,258), Trepagnier (US 2007/0219720), and McNaughton et al. (2011) supply roadway/environment context for mode decisions and motion planning. Motivation: if a narrower or more geometrically demanding roadway reduces the margin for error, it is a predictable design choice to let it shift the assignment toward whichever operator is more competent there. Strong ground — this limitation is well covered.

---

12. Where the § 103 theory is actually vulnerable (do not skip this)

1. E6/E7 — the head-to-head comparison. As drafted (per the family's language), the claims assign control "based on which of the control processor competence level and the human driver competence level is relatively higher to one another." The art as I have it compares each level to a threshold, not to the other level. Tesla's own petition papered over this by arguing it would have been obvious to "add (or substitute) an additional scalar for comparison against Attard's Φ." That is a threshold comparison, not a two-score comparison, and a Patent Owner will say so. Remedies: (a) argue that "relatively higher" reads on normalized scores evaluated against a common scale; (b) add art that blends driver and vehicle inputs into a single comparable index — An's risk index is the best available candidate; (c) add Coelingh / DeRuyck / Gunderson / Cullinane for driver-vs-system arbitration. This is the single most important thing to shore up, and it is claim-construction-dependent.
2. "Driver profile" as a received data structure. The '986 describes a retrieved driver profile (traits, license restrictions such as a night-driving restriction). Hampiholi derives driver state in real time; that is not the same as a stored profile. Attard's operator-associated parameter 117 helps, but a strict construction ("receive a driver profile") may require stored-profile art. Weak-to-moderate.
3. Cohort / active learning (D7) — as discussed in § 9. Weakest.
4. Claim text is unverified. If the '986 claims differ from the family language, everything above must be re-mapped. Foundational risk.

---

13. Anticipation watchlist (do not plead as § 103 without checking)

Some of the family's art may anticipate a broader claim rather than render it obvious — and anticipation is the cleaner case:

• Cullinane — US 8,825,258 (Google/Waymo): transitioning from manual to autonomous based on environmental, vehicle, and driver data. Potentially anticipatory of a claim lacking the explicit "competence comparison."
• Lind et al. — US 9,463,793 (Volvo; EP priority 2014-04-08): method for transition between autonomous, partly-autonomous, and manual modes. Potentially anticipatory of E3/E7 in a bare form.
• Tsimhoni-323 — US 9,798,323 and Sako — US 9,950,740: handover/driver-monitoring. Check for full E1–E7 coverage.

---

14. Secondary considerations (objective indicia)

I found no evidence of objective indicia supporting non-obviousness, and there is an affirmative reason to expect none: the patentee has never sold a product embodying the claims (IBM is an enterprise-IT company; the chain is NPE-only), so there is no nexus between any commercial success and these claims. The only transactional "success" is the reported $9M bulk portfolio divestiture of 500+ US assets — a portfolio sale, not evidence of demand for this invention, and courts routinely refuse to credit such sales absent a nexus. Likewise, no industry praise, no copying, no long-felt-but-unmet need has been evidenced. If Patent Owner later asserts licensing success, test it for a nexus to the claimed handover logic. Absence of objective indicia materially strengthens every ground above.

---

15. Bottom line

• The core of the '986 is obvious over Attard + Hampiholi, with Attard supplying the on-board processor's competence assessment and mode decision and Hampiholi supplying the driver-competence measure for the same operating condition — combined on the strength of Attard's own operator-associated parameter 117, the complementary halves of a single well-known handover problem, structural/hardware compatibility, and the predictable-result and known-technique-improving-a-similar-device rationales.
• The dependent-grade features layer on cleanly: Frazer (fault-remediation table), McNew/Scofield (actual handover, alerts, driver-state sensing), An/Kang (weighted voting and weighted sums), Dannheim/Harding/Muldoon/Caveney/Stenneth (weather, V2V, and vehicle fault sensing), Cullinane/Trepagnier/McNaughton (roadway geometry and width).
• Two soft spots: (i) the head-to-head comparison in E6/E7, which the threshold-based art does not literally reach without claim-construction support or An; and (ii) the cohort / active-learning feature, which is the family's admitted narrow point (see the '765 disclaimer).
• Nothing is adjudicated. Five petitions against the family were denied institution on discretionary grounds; the Federal Circuit denied mandamus on 2026-02-27; the cert petition Tesla, Inc. v. Granite Vehicle Ventures LLC, No. 26-136 is pending. So this section is the only thing standing between the claims and a merits ruling — and the § 315(b) clock and the current discretionary-denial posture are the real obstacles, not the art.

---

Confidence statement

• High confidence: the substance and teaching of Attard (US 9,406,177) and Hampiholi (US 2016/0267335), including Hampiholi's 2015-03-13 filing date resolving the § 102 date issue; the identities of Frazer (US 9,494,926), McNew (US 10,377,303), An (US 9,063,543), and the secondary art set; and the legal framework.
• Moderate confidence: the precise element-by-element mapping of the E1–E7 core onto Attard + Hampiholi, and the strength of the "cohort/active learning" and "driver profile" grounds, both of which turn on claim construction I cannot verify.
• Low confidence / unverified: the verbatim granted claims of US 9,566,986, and the effective filing dates of Frazer, McNew, An, DeRuyck, Gunderson, Coelingh, Hada, Sako, Duncan, Nagasaka, and Schunder — each must be independently confirmed to predate 2015-09-25 before any petition or invalidity contention relies on it. I have not asserted a mapping I could not ground in a retrieved source, and I have flagged rather than fabricated the claim text.

Sources relied on: Attard, US 9,406,177 B2 (patentimages PDF, including claim and ¶ text); Hampiholi, US 2016/0267335 A1 (patentimages PDF and freepatentsonline text); Tesla's IPR2025-00943 petition and grounds summary (ai-lab.exparte.com); the IPR2025-00944 and IPR2025-01034 petition/declaration excerpts (ptacts.uspto.gov and ai-lab.exparte.com); the IPR2025-01034 case summary; Lind et al., US 9,463,793; and the SAE J3016 / Meyer / Harding / Muldoon / Dannheim / Bimbraw / Mosquet citations as listed in the family's exhibit tables.