US 10866101

I'll verify the patent's bibliographic data and check for any litigation or assignment activity against live sources, then summarize.

Let me check for any litigation involving this patent or its assignee, and confirm the current assignee status.

Verification note

I searched for the literal identifier 10866101 / US 10,866,101 B2 against Google Patents, Justia, the patent PDF, Unified Patents' portal, the Federal Circuit docket coverage for 2026, and district/state court dockets. Findings below are drawn from the authoritative granted patent plus live bibliographic records; where third-party aggregators conflict with the patent document, I flag it rather than silently reconciling it.

1. Bibliographic data (US 10,866,101 B2)

Assignee caveat: the chain recorded on Google Patents is TuSimple → TUSIMPLE, INC. → CreateAI, Inc. (change of name, 2025-12-03). A corresponding Australian register entry shows "TuSimple, Inc." altered to "CreateAI, Inc." I did not independently verify the USPTO assignment reel/ frame, so treat the current-owner line as high-confidence-but-not-verified-against-the-native-UPSTO-record.

Third-party data conflict (flagged, not auto-corrected): the Unified Patents portal page for US-20180357503-A1 lists a priority date of 2017-06-12 and a "Grant Date" of 2018-12-12. Both conflict with the authoritative documents (filed/priority 2017-06-13; granted 2020-12-15; the 2018-12-12 date is in fact the publication date). I treat the portal entries as erroneous aggregator data.

2. Abstract (verbatim)

"A system for generating a ground truth dataset for motion planning is disclosed. The system includes sensors for data acquisition for motion planning of a vehicle, the sensors including a LiDAR; a calculating module configured to calculate LiDAR poses in response to data from the sensors and generate undistorted LiDAR scans; an extracting module configured to extract static points from a target undistorted LiDAR scan and generate a LiDAR static-scene point cloud; and a generating module configured to generate sparse image point correspondences for each pair of images, using the LiDAR static-scene point cloud."

3. Plain-language overview of the independent claims

Claim 1 — System for motion planning. A system with a memory, a sensor set including a LiDAR sensor, and a processor that:

• performs a LiDAR scan resulting in a pair of images, each image containing observed points and each taken at a different time (note: the claim literally says the LiDAR scan results in the image pair — an artifact of the drafting; the specification describes camera images, not LiDAR-produced images);
• generates a LiDAR static-scene point cloud by extracting static points using a transformed reference scan; and
• generates an image point correspondence for the pair of images using that static-scene point cloud.
  The processor is further required to (i) identify static points belonging to static objects for the pair of images based on a comparison between a point from the LiDAR scan and a point in the transformed reference scan, and (ii) align the observed points based on that identification of static points.

In substance: use motion-corrected LiDAR to decide which sensed points belong to stationary world structure, then use only those stationary points to establish correspondences between two images taken at different times — i.e., sparse "ground truth" optical flow for a static scene.

Claim 14 — Non-transitory computer readable storage medium. Substantively parallel to claim 1, reciting the same three operations (perform LiDAR scan yielding a pair of images at different times; generate the static-scene point cloud via a transformed reference scan; generate an image point correspondence) and the same static-point identification/alignment requirement.

⚠️ Claim-drafting irregularity (noted literally, not corrected): claim 14 is a method-type claim ("cause the computing device to implement a method comprising…"), yet its wherein-clause recites "wherein the at least one processor is further configured to…" — processor-apparatus language inside a method claim, and a reference to "the at least one processor" that has no antecedent in that claim. This is a real, citable defect in the granted text and is a likely source of §112(b) or indefiniteness argument if the claim is ever asserted.

Dependent claims in brief

• 2 — sensors further include a camera and a GNSS-inertia system; LiDAR scan performed in "swing" form.
• 3 — filter out image point correspondences related to invalid pixels.
• 4 — detect a GNSS-inertial pose estimate near the acquisition time of the LiDAR scan and align it with the observed points.
• 5 — determine LiDAR pose by interpolation; transform the LiDAR scan into an undistorted scan.
• 6 — determining the LiDAR pose includes interpolating relative positions from a GNSS-inertial system.
• 7 — static points identified using interpolation of relative positions; observed points aligned based on GNSS-inertial estimates.
• 8 — the transformed reference scan is temporally proximate to the LiDAR scan and aligned with it.
• 9 — distance between a static point from the LiDAR scan and a point in the transformed reference scan is below a threshold, and the static point is extracted.
• 10 — correspond the first image of the pair to the LiDAR static-scene point cloud.
• 11 — project a 3D point onto the pair of images to yield two image points in sub-pixel accuracy, then establish the correspondence.
• 12 — compare local image patches near correspondences and filter out those with a significant patch difference (occlusion rejection).
• 13 — store filtered correspondences as ground truth for a static-scene dataset motion planning.
• 15–20 — mirror claims 3–9 in medium form.

Terminology inconsistencies worth noting (relevant to claim construction): the specification alternates between "sweep" and "swing" for a 360° LiDAR rotation; "GNSS-inertial" vs. the recurring typo "GNSS-insertion"; and "accusation time" where "acquisition time" is meant (claims 4/16 use the correct word). The specification's own disclosure is at times broader than the claims (e.g., it describes stitching scans into a local map and ICP refinement at FIG. 1 operations 108/110, which do not appear in the independent claims).

4. Family and related filings

The specification incorporates by reference and claims priority to seven co-pending applications filed the same day (June 13, 2017), same inventors — including US 10,488,521 ("Sensor Calibration and Time Method…"), US 10,481,267 ("Undistorted Raw LiDAR Scans and Static Point Extractions Method…"), and US 10,360,686 ("Sparse Image Point Correspondences Generation and Correspondences Refinement System…"), plus four applications that published (e.g., US 2018/0356824, US 2018/0357314, US 2018/0357315, US 2018/0356831) and are now abandoned. Only US 10,866,101 is the subject of this query.

5. Docket / litigation check (as of April 26, 2026)

• CAFC 2026 dockets: I found no appeal, IPR appeal, or other Federal Circuit proceeding involving US 10,866,101. I reviewed the Federal Circuit weekly summaries for the year to date (e.g., weeks ending Feb. 13, 2026 and Apr. 3, 2026) and the Q Technologies v. Walmart opinion docket coverage; none references this patent. Stated with appropriate uncertainty: CAFC coverage summaries are not a complete docket substitute, so this is a negative finding from available search results, not a certified docket search.
• Extant litigation involving the assignee family (relevant context, but not a suit asserting this patent): TuSimple, Inc. v. Bot Auto, Inc. (a/k/a CreateAI Inc v. Bot Auto Inc), S.D. Tex. No. 4:25-cv-03371, filed July 21, 2025, Judge Charles Eskridge — docketed as a patent-infringement case, with a motion to dismiss and a motion to stay pending as of Jan. 16, 2026. From the available complaint analysis, the identified asserted patents are US 11,577,776, US 12,190,609, and a "…'783" patent — I did not locate 10,866,101 among the asserted patents, so do not attribute this case to the '101 patent without confirming the complaint's exhibit list.
• State proceedings: parallel Texas trade-secret litigation, CreateAI Holdings, Inc. (f/k/a TuSimple Holdings, Inc.) v. Bot Auto TX Inc., No. 15-25-00001-CV (15th Court of Appeals; interlocutory appeal from Harris County Business Court No. 24-BC11A-0007), resulted in affirmance of the denial of a temporary injunction (memorandum opinion filed September 15, 2026). Also, a derivative/investor action, Wilhoite v. Hou, S.D. Cal. No. 23-cv-2333, settled for $42.5M with final approval July 23, 2025. All of this is corporate/technology context, not patent-validity or infringement adjudication of the '101 patent.

6. Confidence and open items

• High confidence: title, inventors, assignee of record (original), filing/priority date (2017-06-13), issue date (2020-12-15), application number, abstract, claim set, PTA-adjusted expiry (2037-08-29), and family relationships — all taken from the granted patent text and PDF.
• Moderate confidence: the Dec. 3, 2025 change-of-name to CreateAI, Inc. (consistent across Google Patents and the Australian register, but I did not read the USPTO assignment document itself).
• Explicitly uncertain / unverified: (a) no litigation asserting this specific patent was found; (b) the Unified Patents portal dates (2017-06-12 priority; 2018-12-12 "grant") are contradicted by the authoritative record and I did not adopt them; (c) I have no authoritative information on any pending inter partes review, reissue, or reexamination of US 10,866,101 beyond the fact that its recorded legal status is "Active."

I'll search for litigation involving US patent 10866101 across patent litigation databases and court sources.

Let me search more specific litigation databases and sources.

Search Results: Litigation Involving U.S. Patent No. 10,866,101

Bottom line: I could not identify any litigation in which U.S. Patent No. 10,866,101 ("the '101 patent") is specifically named as an asserted patent. I searched patent-litigation aggregators and docket sources (Unified Patents, Justia Dockets, CourtListener, USPTO assignment/PTAB records) using the exact number "10866101" and its variants. No case returned that asserts this specific patent.

Below is the most relevant related litigation I found involving the patent's owner, with the important caveat about the '101 patent itself.

1. TuSimple, Inc. v. Bot Auto, Inc. et al. (the likely relevant case)

Critical caveat: The public complaint analysis for 4:25-cv-03371 identifies patents such as U.S. Patent No. 11,577,776 and U.S. Patent No. 12,190,609 (map construction), among others in TuSimple's portfolio. I did not find any source confirming that U.S. Patent No. 10,866,101 is asserted in this complaint. Because TuSimple asserted a multi-patent portfolio, the '101 patent could be among the asserted patents, but I cannot state that with confidence based on the available results, and you should verify directly against the complaint (Dkt. 1) and its exhibits in PACER.

2. Other matters involving the same parties (NOT patent litigation over the '101 patent)

• CreateAI Holdings, Inc. / TuSimple, Inc. v. Xiaodi Hou, Lei Wang, Haoran Wang, Yujia Wu, Qian Chu (and Bot Auto) — California Superior Court, County of San Diego; complaint filed March 21, 2025. This is a trade-secret / breach-of-contract action (misappropriation of AV technology, sensor designs, AI models), not a patent-infringement action, and it does not assert the '101 patent.
• Dicker v. TuSimple Holdings, Inc., No. 3:22-cv-01300 (S.D. Cal.) — securities class action, not patent litigation.

3. PTAB / post-grant proceedings

I found no IPR, PGR, or other post-grant challenge naming U.S. Patent No. 10,866,101. (Unified Patents' PTAB filings I reviewed involved unrelated patents.)

Summary

• Known litigation specifically asserting U.S. 10,866,101: none identified.
• Closest related matter: TuSimple, Inc. v. Bot Auto, Inc. et al., No. 4:25-cv-03371 (S.D. Tex., filed July 21, 2025) — a live patent-infringement suit by the '101 patent's owner against a competitor. Whether the '101 patent is one of the asserted patents requires direct verification against the complaint and its exhibits, which I was not able to confirm from the search results obtained.

Recommendation: To definitively confirm or exclude the '101 patent, pull the Docket 1 complaint and its patent exhibits in 4:25-cv-03371 via PACER, and run the patent number through patent-litigation trackers (RPX/Unified Docket, Lex Machina) — note that the current assignee is CreateAI, Inc. (renamed from TuSimple, Inc., effective Dec. 3, 2025), so filings may appear under either name.

If you would like, I can attempt a more targeted search of the 4:25-cv-03371 complaint exhibits to confirm whether the '101 patent is listed.

I'll verify the structured block's default by searching for any PTAB activity on US 10,866,101.

Let me run a few more targeted checks before concluding.

PTAB / AIA Trial Activity — US 10,866,101 B2

Patent: US 10,866,101 B2 — "Sensor calibration and time system for ground truth static scene sparse flow generation"
Appl. No.: 15/621,884 · Filed: 2017-06-13 · Granted: 2020-12-15
Inventors: Yi Wang, Bolun Zhang, Yi Luo, Ke Xu
Original assignee: TuSimple, Inc. → Current assignee: CreateAI Inc (change of name, recorded 2025-12-03)
Adjusted expiration: 2037-08-29 · Claims: 1–13 (system), 14–20 (non-transitory CRM)

Proceedings overview

Total AIA trial proceedings on file: 0. The structured PTAB block supplied with this prompt — sourced from the USPTO Open Data Portal — returns no AIA trial proceedings (no IPR, no PGR, no CBM) for US 10,866,101, and my independent web searches surfaced no petition, institution decision, Final Written Decision, or Federal Circuit appeal involving this patent number. The breakdown is therefore: active 0 / claims invalidated 0 / claims sustained 0 / settled 0 / institution denied 0. The bottom line for a defendant is the opposite of the usual "patent is hardened" story: this patent is untested at the PTAB, which means there is no § 315(e)(2) estoppel wall in your way and no petitioner-generated record to lean on. Every claim — including independent claims 1 and 14 — is live and presumed valid under § 282. The absence of PTAB activity on a 2020-issued patent in a hot art (LiDAR/camera calibration for AV ground truth) is a signal worth reading carefully: this patent's family appears to have been maintained as portfolio/prosecution assets rather than as an assertion vehicle that attracted third-party invalidity challenges.

Per-proceeding detail

No proceedings to report. I am deliberately not generating placeholder sections, because doing so would require inventing proceeding numbers, which is prohibited by the source instructions and by the anti-fabrication constraint.

For the record, here is what I checked and what came back empty:

The only hits for "10,866,101" outside the patent itself were unrelated (a GenBank GI number, an unrelated real-property ESA report). None are PTAB proceedings.

Strategic summary

Claim status — all 20 claims are UNTESTED and INTACT. No claim of US 10,866,101 has been canceled, narrowed, disclaimed, or held unpatentable in any AIA trial. That includes:

• Independent claim 1 (system: sensors incl. LiDAR; processor to perform a LiDAR scan resulting in a pair of images, generate a LiDAR static-scene point cloud using a transformed reference scan, generate an image point correspondence, and identify/align static points based on a comparison between a point from the scan and a point in the transformed reference scan).
• Independent claim 14 (the same subject matter recast as a non-transitory CRM).
• Dependent claims 2–13 and 15–20 (camera + GNSS-inertia system, invalid-pixel filtering, pose interpolation/undistortion, thresholded closest-point extraction, sub-pixel projection, patch-comparison refinement, storing filtered correspondences as ground truth).

Estoppel landscape — a clean slate. Because no petitioner has ever filed an AIA petition against this patent, no § 315(e)(2) estoppel attaches to anyone. Any defendant is free to raise any ground, including § 102/§ 103 art that was before the examiner during prosecution (subject only to the § 325(d) discretion and § 315(a)(1) "filed-before-civil-action" bars, neither of which applies yet). There is also no IPR-driven narrowing that could give the patent owner a "we already survived the art" narrative. On the other hand, there is no prior PTAB record to mine for admissions, claim constructions, or expert testimony — you would be building the invalidity case from scratch. If you are served with a complaint, note the § 315(b) clock: one year from service to file an IPR.

Pattern signals. The patent is part of a large, commonly-owned TuSimple/CreateAI family (the specification incorporates eight co-pending Jun. 13, 2017 applications by reference, several of which matured into US 10,488,521, US 10,481,267, and US 10,360,686). Despite that portfolio, this particular patent shows no litigation-driven IPR activity, no serial petitioning, and no patent-owner PTAB appeal activity — the owner has not had to defend it at the Board. There is no defensive aggregator in the chain. The Chinese counterpart, CN109087338B ("A method and device for extracting sparse optical flow from images"), likewise shows no challenge activity in the surfaced record.

Recommended next steps

1. Verify the null result at the primary sources before relying on it. The ODP digest is the canonical list here, but it can lag. Re-run the check directly:

   • USPTO PTAB E2E / PTAB Center (https://e2e.uspto.gov/) — search by patent 10,866,101 and by application 15/621,884.
   • USPTO PatentCenter (https://patentcenter.uspto.gov/) — the "Reexams" and "Trials" tabs are authoritative for administrative proceedings (a reexam order, though not an AIA trial, is likewise your problem and would not appear as an "IPR").
   • The patent's Google Patents page (https://patents.google.com/patent/US10866101B2/en) — its Legal Events timeline shows assignment records only (TuSimple → CreateAI change of name, 2025-12-03), no IPR events.
   • Docket Alarm / CourtListener PTAB docket (https://www.courtlistener.com/) for any recently-filed 2025–2026 petition not yet propagated to the ODP.

2. If a demand letter or complaint cites US 10,866,101, do not assume it is battle-tested. It is not. The absence of any PTAB proceeding is the key defensive datum: there is no FWD to quote, no cancellation to leverage, and no estoppel to invoke. Your invalidity posture is "clean-sheet IPR," filed within one year of service under § 315(b).

3. If you want to be the first mover, an IPR against claim 1 would be an unconstrained attack — full range of § 102/§ 103 art, no § 325(d) drag from prior AIA proceedings, and the patent's own specification helpfully enumerates the prior-art landscape it builds on (Levinson & Thrun, "Unsupervised Calibration for Multi-beam Lasers"; Zhang & Singh, "LOAM: Lidar Odometry and Mapping in Real-time"; Geiger et al., "Automatic Camera and Range Sensor Calibration Using a Single Shot"; Z. Zhang, "A Flexible New Technique for Camera Calibration"; the KITTI benchmark literature). Those admissions are fair game.

Caveat: I found no evidence of any PTAB proceeding, and I have not fabricated one. If a proceeding exists that I could not surface, it is not in the ODP ingest and did not appear in any indexed public source at the time of this analysis — treat the "zero proceedings" finding as strong but subject to the direct-source verification in step 1.

I'll verify the ownership events and corporate history against live sources before reconstructing the chain.

Now let me check for litigation naming this patent, assignment reel/frame records, and the inventors' affiliations.

Ownership & Assignment Reconstruction — US 10866101 B2

Patent: US 10866101 B2 — "Sensor calibration and time system for ground truth static scene sparse flow generation"
App. No.: 15/621,884 · Filed: 2017-06-13 · Issued: 2020-12-15 · Legal status: Active, adjusted expiration 2037-08-29
Family: one of eight co-pending applications filed 2017-06-13 by the same four inventors (this one is the "system" sibling of US 10,488,521; the family also produced US 10,481,267 and US 10,360,686, with four siblings abandoned). CN family member: CN 109087338 B, filed 2018-02-26 claiming the same 2017-06-13 priority.

Inventors

Employer determination: all four are named as assignors in the 2017-06-13 assignment of assignors' interest to TuSimple recorded on the filing date itself, which is the standard footprint of an employee invention-assignment (EPIIA) executed at or before filing. TuSimple was headquartered in San Diego, CA with a large Beijing/Shanghai engineering operation; the inventors' work location is not stated in the public record.

Unusual patterns: none tied to the inventors individually. No evidence any of the four departed within 12 months of filing — the opposite, in fact, as the same four re-appear across all eight sibling filings. The notable later event is company-wide, not inventor-specific: TuSimple's December 2023 layoff round, the elimination of the algorithm team, and a declaration in the Bot Auto litigation that "nearly 100% of the AI engineers have left TuSimple before June 2024." That is a mass exodus from the organization, not a precursor fire-sale signal for this patent.

Original assignee

TuSimple (name of record at filing; renamed TUSIMPLE, INC. by 2020; now CreateAI, Inc. as reflected in Google Patents' "Current Assignee").

• Business: autonomous trucking / L4 autonomous driving technology (perception, sensor fusion, mapping, motion planning). Founded 2015; went public on Nasdaq (TSP) in April 2021, the first autonomous-driving company to do so, raising ~$1.35B at an ~$8.5B valuation.
• Did it ship a product embodying the claims? Yes, during the relevant period. TuSimple built and operated an autonomous truck platform, ran driver-out hub-to-hub runs, and had Navistar/Traton commercial partnerships. The claimed subject matter — sensor calibration, time synchronization and ground-truth static-scene sparse flow generation from LiDAR + camera + GNSS/IMU — is squarely the internal data-annotation pipeline that fed its perception stack (the litigation record describes five CNN models for camera, lidar, fusion, tracking and prediction, all trained on manually annotated road data).
• Current status: Operating, but no longer in the field of the claims. US truck operations wound down (announced Dec 2023); delisted from Nasdaq (Form 25 ~Jan 29, 2024; shares ceased trading ~Feb 7, 2024). Rebranded CreateAI on 2024-12-18/19, pivoting to generative AI for animation and video games. Delaware parent TuSimple Holdings, Inc. → CreateAI Holdings, Inc. effective 2024-12-11; the operating subsidiary TuSimple, Inc. → CreateAI, Inc. is the recorded owner of this patent as of the 2025-12-03 recording. The company is in active shareholder litigation (Delaware Chancery and S.D. Cal.) over the transfer of its remaining ~$450M cash to China and possible self-dealing by co-founder Mo Chen; former CEO Xiaodi Hou is publicly demanding liquidation. No Chapter 7 or Chapter 11 filing was identified in the sources reviewed.

Assignment timeline

Data limitation — read first. The authoritative source for this section is USPTO Assignment Center. The records exposed to me via the Google Patents legal-events mirror do not display reel/frame numbers, execution dates, or the correspondent of record for any of the three entries. I have therefore not invented reel/frame identifiers. Items marked not exposed must be pulled manually at the Assignment Center search page (link in the Verdict section). None of the three records is an assignment to a third party; all three stay inside one corporate lineage.

1. 2017-06-13 (executed and recorded, filing date) — Reel/frame: not exposed

• Conveyance: Assignment — "Assignment of Assignors' Interest (see document for details)"
• Assignor: YI LUO; YI WANG; KE XU; BOLUN ZHANG (the four named inventors)
• Assignee: TuSimple
• Correspondent: not exposed
• Context: Original employee invention assignment; all four inventors conveyed their rights to the applicant on the filing date.

2. Execution date not exposed / recorded 2020-01-30 — Reel/frame: not exposed

• Conveyance: Change of Name
• Assignor: TuSimple
• Assignee: TUSIMPLE, INC.
• Correspondent: not exposed
• Context: Internal corporate name change only — no change in beneficial ownership, no consideration, no third party.

3. Execution date not exposed / recorded 2025-12-03 — Reel/frame: not exposed

• Conveyance: Change of Name
• Assignor: TUSIMPLE, INC.
• Assignee: CreateAI, Inc.
• Correspondent: not exposed
• Context: Internal name change only, reflecting the group's December 2024 rebrand. Independently corroborated: the Delaware certificate amendment renaming the parent (TuSimple Holdings → CreateAI Holdings) was filed and approved 2024-12-11, and plaintiff's counsel filed a "Notice of Name Change by Subsidiary TuSimple Inc." in TuSimple, Inc. v. Bot Auto, Inc., No. 4:25-cv-03371 (S.D. Tex.) on 2025-10-06.

No fourth record exists. There is no recorded assignment of US 10866101 out of the original corporate family.

Timeline diagram

timeline
    title Ownership of US 10866101
    2017 : Inventors assign to TuSimple on filing date
    2020 : TuSimple renamed TUSIMPLE INC
    2021 : IPO on Nasdaq as TSP
    2023 : US autonomous truck programme wound down
    2024 : Delisted from Nasdaq
         : Parent renamed CreateAI Holdings
    2025 : TuSimple Inc renamed CreateAI Inc
         : Sibling patents asserted v Bot Auto

NPE / troll-pattern signals

1. Shell-entity transfer — NOT PRESENT.
The chain runs inventors → TuSimple → TUSIMPLE, INC. → CreateAI, Inc. Each link is the same Delaware entity lineage, evidenced by two "Change of Name" conveyances (recorded 2020-01-30 and 2025-12-03) rather than an assignment for value. No "IP / Licensing / Patents / Ventures"-suffixed vehicle ever took title to this patent, and there is no single-purpose LLC anywhere in the chain. The "CreateAI Holdings" name refers to the public-reporting parent, not to the recorded assignee.

2. Known asserter in the chain — NOT PRESENT.
None of TuSimple, TUSIMPLE, INC., or CreateAI, Inc. matches Acacia, Marathon Patent Group, Intellectual Ventures, IPNav, Wi-LAN, Mosaid/Conversant, Vringo, Pendrell, Innovatio, MPHJ, Lumen View, Round Rock, Document Generation Corp, or an Erich Spangenberg entity. Caveat: I could not pull RPX / Unified Patents asserter directories in this pass, so this is a negative finding against the named lists only, not a clean bill of health against every high-frequency-plaintiff index.

3. Repeat correspondent across the chain — UNCLEAR / NOT ASSESSABLE.
This is the one signal I could not test, and it is the highest-value item to obtain. The Google Patents legal-events mirror does not expose the correspondent of record for any of the three recorded events, so I cannot say whether one attorney or firm filed both the 2020-01-30 and 2025-12-03 name-change recordings. Action item: retrieve the correspondent field for both records from Assignment Center. Note that both recordings are Change of Name (not enforcement-adjacent) conveyances, so recurrence here — even if found — would carry far less weight than recurrence across an assignment-to-asserter chain.

4. Cascading transfers — NOT PRESENT.
Two conveyances in eight and a half years, both name changes, zero transfers through chained LLCs, zero assignees sharing a correspondent address. The classic "<24 months through three shells" footprint is absent.

5. Pre-litigation transfer — NOT PRESENT (and the timing is inverted).
No assignment of US 10866101 occurred within six months before any infringement action; in fact no assignment of this patent has ever occurred beyond the original grant from the inventors. More importantly, I found no pleading asserting US 10866101. The July 21, 2025 S.D. Tex. patent case (TuSimple, Inc. v. Bot Auto, Inc. et al, 4:25-cv-03371, Judge Charles Eskridge) asserts a different set of family-adjacent patents — a sensor-configuration patent ("'783", forward/backward camera focal-length limitation), US 11,577,776 (redundant steering), and US 12,190,609 (map construction) — per a third-party complaint analysis I reviewed (not the complaint itself, which I did not read). Separately, the 2025-12-03 name-change recording postdates that lawsuit, which is the reverse of the arranged-chain pattern.

6. Bankruptcy fire-sale — NOT PRESENT as a matter of record.
No Chapter 7 or Chapter 11 filing, and no §363 sale, was identified. The distress is real and documentable — delisting, US wind-down, ~$450M cash subject to a shareholder fight, a rejected unsolicited proposal from Steel Partners and Camac Partners (2024-12-31), and a public liquidation campaign by co-founder Xiaodi Hou — but distress is not a bankruptcy sale, and no patent has moved as a result. Watch item: if a liquidating Chapter 11 or an ABC is later filed, this portfolio is a plausible sale asset.

7. Privateering — NOT PRESENT.
The inverse is true: the operating entity's own corporate family is the plaintiff, suing a direct competitor (Bot Auto, founded by its former co-founder/CEO and staffed by former TuSimple personnel) under the Texas Uniform Trade Secrets Act and, separately, in federal patent court. No NPE was interposed to assert on TuSimple/CreateAI's behalf.

8. Defensive aggregator — NOT PRESENT.
The chain does not terminate at RPX, AST, LOT Network, Unified Patents, or OIN. Title remains with the original operating lineage, so the patent has not been neutralized.

One forward-looking caveat, explicitly not a finding: CreateAI's published business plan references monetizing its existing autonomous-driving IP, and the company has publicly said it intends to license its self-driving technology to partners in China. Nothing in the recorded assignment chain shows an intent to transfer to an assertion vehicle, and today's holdings are consistent with ordinary corporate housekeeping. This is flagged only as an item to re-check if a future recording names a new, non-operating assignee.

Verdict

INSUFFICIENT DATA — more precisely: no NPE pattern is present on the current record.

Three assignment records exist, and none transfers ownership to a third party: the original 2017-06-13 assignment of the four inventors' interest to TuSimple, and two Change of Name conveyances (recorded 2020-01-30, TuSimple → TUSIMPLE, INC.; recorded 2025-12-03, TUSIMPLE, INC. → CreateAI, Inc.). Title never left the original corporate lineage, so zero of the eight NPE signals is affirmatively present. The verdict is deliberately not "Operating-company assertion" despite the family's 2025 patent suit against Bot Auto, for two reasons: (a) that suit asserts sibling patents, not US 10866101, and (b) the current assignee has exited the field of the claims entirely (US autonomous trucking wound down 2023–24, AV R&D ended, pivot to generative AI animation/gaming in Dec 2024), so it does not "ship a product embodying the claims" as that verdict requires. The residual uncertainty is narrow but material: correspondent-of-record and reel/frame data were not exposed in the fetched records, and RPX/Unified asserter directories were not consulted.

Verify at:

• USPTO Assignment Center — https://assignmentcenter.uspto.gov/ (search "10866101", then retrieve reel/frame and the correspondent field for the 2020-01-30 and 2025-12-03 recordings)
• USPTO Assignment Search mirror — https://assignment.uspto.gov/patent/index.html
• Google Patents legal events (source of the three entries above) — https://patents.google.com/patent/US10866101/en
• Docket corroboration — TuSimple, Inc. v. Bot Auto, Inc. et al, No. 4:25-cv-03371 (S.D. Tex.), https://dockets.justia.com/docket/texas/txsdce/4:2025cv03371/[2019667](/patent/2019667)

I'll search for the USPTO record of US 10,866,101 and its cited references.

Let me retrieve the patent's front-page references and the Google Patents citation tables directly.

Let me pull the citation tables from the publication and sibling pages.

Prior-Art Analysis — US 10,866,101 B2 (application 15/621,884)

Identifier discipline: the requested number 10866101 resolves, literally and uniquely, to US 10,866,101 B2 — "Sensor calibration and time system for ground truth static scene sparse flow generation," app. 15/621,884, filed 2017‑06‑13, granted 2020‑12‑15 (see the patent summary already generated; not repeated here). I did not substitute or auto-correct any similar-looking number. Search results that could be confused with the ID (e.g., a Helicobacter pylori GenBank GI number "10866101" surfaced in one hit) are irrelevant and were discarded.

⚠️ Important scope limitation — read before using the tables

The task asks for "each patent citation for 10866101." I must be candid about what I could and could not retrieve:

A useful lead for the missing table: the sibling method patent US 10,488,521 B2 ("Sensor calibration and time method…," same family, same inventors, filed the same day) is reported by Google Patents to have a "Patent Citations (17)" table — see the snippet at https://patents.google.com/patent/[US10488521B2](/patent/US10488521B2)/en#5. Because the '521 and '101 disclosures are substantially identical, the examiner's cited U.S. patent documents for '521 are a strong proxy for those on '101 — but that is an inference, not a verification, and I flag it as an open item rather than presenting a proxy list as fact. To close it, pull the front page of the '101 grant (PDF: https://patentimages.storage.googleapis.com/f8/be/25/d5bbb3ebbda644/US10866101.pdf) or the "Patent Citations" section at https://patents.google.com/patent/[US10866101B2](/patent/US10866101B2)/en.

1. Legal framework applied

• The '101 patent has an effective filing date of 2017‑06‑13 (no earlier foreign/or domestic priority; the CN case CN109087338B claims priority to this US filing, not the reverse — see the earlier summary). It is therefore a post‑AIA (first‑inventor‑to‑file) patent.
• Prior art is assessed under 35 U.S.C. §102(a)(1) (patented, described in a printed publication, or otherwise publicly available before 2017‑06‑13) and §102(a)(2) (U.S. patents/published applications effectively filed before that date).
• Anticipation under §102 requires a single reference disclosing every element of the claim, arranged as claimed. Where a reference discloses only some elements, it is §103 (obviousness) material, not §102 art. I note this distinction for each entry.
• All seven specification references below were published between 2000 and 2014 — all qualify as §102(a)(1) printed publications. None, standing alone, discloses the full combination of independent claim 1 or claim 14.

2. References cited in US 10,866,101 (verified from the specification)

The '101 specification expressly lists the following as references incorporated by reference (source: the patent's "All referenced patents, applications and literatures throughout this disclosure are incorporated herein by reference…" paragraph, confirmed at https://www.patents-review.com/a/20180357503-sensor-calibration-time-system-ground-truth-static-scene.html).

Bottom line on the verified cited references: they are the building blocks the '101 specification itself leans on (calibration per R1/R3/R5, motion-distortion correction per R2, benchmarking per R4/R6/R7). Individually and even collectively they describe the inputs and evaluation of a ground-truth pipeline but not the claimed method of deriving static points from a transformed reference scan and projecting them to obtain sparse image-point correspondences. None of R1–R7 anticipates claims 1 or 14 under §102; their realistic use is a §103 obviousness attack, and even that would need a rationale to combine a calibration paper (R1/R3), an odometry paper (R2), and a benchmarking paper (R4/R6/R7) to reach the claimed pipeline — a combination the prior art does not suggest on its face.

3. Family references — expressly not §102 prior art

The specification incorporates by reference seven same‑day applications (all filed 2017‑06‑13, same inventors, same assignee): US 10,488,521; US 10,481,267; US 10,360,686; and four now‑abandoned cases (US 2018/0356824, US 2018/0357314, US 2018/0357315, US 2018/0356831). Because they share the '101 filing date and inventive entity and were commonly owned, they are excluded from prior art by the §102(b)(2)(C) common‑ownership exception (and, for the applications, by co‑pending/same‑date considerations). They should not be cited as §102 art against '101. (Cross-reference: the family relationships were set out in the earlier "Family and related filings" section — not repeated here.)

4. Forward citations — also not §102 art (they post-date the '101 patent)

For completeness, the only "Cited By" entries on the '101 record are:

Related forward citations in the "Families Citing this family" table (e.g., EP 4 001 965 A1, Volvo Truck Corp., "Lidar localization using optical flow," priority 2020‑11‑17) likewise post‑date the '101 patent and are irrelevant as §102 art.

5. Anticipation conclusions (per independent claim)

• Claim 1 (system): No reference located discloses, in a single document, all of: (a) a sensor set including a LiDAR performing a scan "resulting in a pair of images" at different times; (b) generating a LiDAR static‑scene point cloud by extracting static points using a transformed reference scan; (c) generating an image point correspondence for the pair using that point cloud; and (d) identifying static points by comparing a point from the LiDAR scan with a point in the transformed reference scan and aligning observed points based on that identification. No §102 anticipation identified among either the specification cites (R1–R7) or the located forward citations.
• Claim 14 (non‑transitory CRM): Same analysis; substantively parallel to claim 1 (its drafting irregularity — "wherein the at least one processor is further configured to…" inside a method‑type claim — was flagged in the earlier summary). No §102 anticipation identified.

6. Confidence and open items (net-new to this section)

• High confidence: that R1–R7 are the references cited in the specification, their bibliographic details, their pre‑2017‑06‑13 publication status, and that none anticipates claims 1/14 (all taken from the authoritative patent text supplied in the record and the patent's own reference paragraph).
• High confidence: that the two "Cited By" documents are forward (post‑dating) citations and cannot be §102 art.
• Explicitly unresolved (the core gap for this task): I was unable to retrieve the examiner‑cited "U.S. Patent Documents" / "Patent Citations" table on the face of US 10,866,101. The requested "each patent citation" list therefore cannot be presented as verified fact from the sources I obtained. Do not treat R1–R7 as a substitute for that table — they are specification references, a distinct category.
  • To close the gap: (1) open https://patents.google.com/patent/US10866101B2/en and read the Patent Citations block; (2) read the grant's front page at https://patentimages.storage.googleapis.com/f8/be/25/d5bbb3ebbda644/US10866101.pdf ("References Cited" → "U.S. Patent Documents"/"Foreign Patent Documents"); (3) pull the USPTO PatentCenter file wrapper for app. 15/621,884 (PTO/SB/08 IDS forms and the examiner's Form PTO‑892 "List of References Cited by Examiner"), which is the definitive §102 citation record. The US 10,488,521 "Patent Citations (17)" table (https://patents.google.com/patent/US10488521B2/en#5) is a strong but unverified proxy given the near-identical disclosure.
• No fabrication: I have deliberately not reconstructed the missing citation table from memory; the operating rules require flagging unknown items over guessing, and auto-correcting or back‑filling an identifier's citation list would violate that.

If you re‑authorize a web lookup, I can pull the '101 "Patent Citations" block and the file‑wrapper IDS/PTO‑892, then produce the per‑reference §102 table (full citation, publication/filing date, description, and claim mapping) for the actual examiner‑cited patent documents — the piece this answer could not verify.

I'll ground this §103 analysis in the actual prior-art references available on this patent's record (the non-patent literature cited in the specification and the earlier-filed family members), verifying their content and dates against live sources.

Strong grounding so far. Now let me verify two additional pillars of the §103 analysis: (a) how the KITTI benchmark generated ground-truth optical flow by projecting LiDAR points into image pairs, and (b) prior art on extracting static/background points from successive LiDAR scans.

Obviousness Analysis — U.S. Patent No. 10,866,101 B2

Under 35 U.S.C. § 103(a); Graham v. John Deere, 383 U.S. 1 (1966); KSR Int'l Co. v. Teleflex Inc., 550 U.S. 398 (2007)

Scope note. This analysis builds on the previously-generated summary, claim-construction section, and litigation section, and does not repeat them. Per the operating rule, all identifiers are read literally; the reference set is drawn from the patent's own record (the specification-cited non-patent literature plus the earlier-filed member of the citing family) rather than from third-party aggregator pages, and forward citations are excluded (see §7).

1. The prior-art set on this record, and its §102 status

Conflicts/notes flagged rather than auto-corrected:

• WO 2019/023892 A1 (DJI, "Correction of motion-based inaccuracy in point clouds") appears in the "Families Citing this family" list but has a 2017-07-31 priority — after the '101 priority date. It is not prior art and I do not use it.
• The seven same-day sibling applications incorporated by reference (US 10,488,521; US 10,481,267; US 10,360,686; and the four abandoned publications) are not §102(a)(2) art (not "effectively filed before") and would in any event be excepted under §102(b)(2)(C). Do not cite them as prior art.

2. Level of ordinary skill

A PHOSITA here has a B.S. (or M.S.) in EE, CS, robotics or a related field plus ~2 years' experience in autonomous-vehicle perception, specifically multi-sensor calibration, point-cloud processing, and image/flow estimation — or equivalent. All seven of R1–R7 are the field's canonical references, and the '101 specification itself names each of them as the starting point for its own sub-steps, which is strong evidence of what a PHOSITA is charged with knowing.

3. Element-by-element mapping of independent claim 1

Every element of claim 1 is disclosed, and the purpose element — sparse LiDAR-derived image correspondence for a static scene — is disclosed by R1.

4. Combination A — R1 + R3 + R4 (+ R2)

Statement of the combination. R1 (KITTI) provides the multi-sensor platform, the calibration pipeline (R2 as its own ICRA-2012 component), the Velodyne-plus-GPS/IMU egomotion compensation, and the LiDAR-projected sparse ground-truth flow. R3 (LOAM) provides the sweep-level pose interpolation, distortion correction to a sweep-start reference time, and the nearest-neighbour + residual-threshold matching used to decide which points are consistent with a reference scan. R4 (Levinson/Thrun) supplies the express assumption that the environment is static, that points lie on contiguous surfaces, and the distance-penalised alignment of points against a reference set. R2 supplies the projected-3D-point-to-image registration at sub-pixel accuracy.

Why a PHOSITA would combine them — articulated reasoning (MPEP 2143/2144.03; KSR):

1. R1 itself supplies the motivation. R1 states the ground-truth flow/stereo density comes from LiDAR points projected into images and that egomotion in the LiDAR must be compensated using GPS/IMU. That is an express teaching to (a) transform LiDAR points into a common reference frame before (b) projecting them into the image pair to make a correspondence — i.e., the whole of claim 1's final step.
2. R1's own accuracy bottleneck points at R3/R4. R1's flow/stereo ground truth is limited by LiDAR-to-camera registration quality and by noise; the '101's solution to that problem — keep only points that are consistent with a transformed reference scan (static), discard the rest — is the predictable, known remedy: R4 states the static-contiguous-surface assumption explicitly, and R3 implements the reference-scan consistency test with a residual threshold. Applying a known reliability filter to a known noisy measurement is "use of a known technique to improve a similar device in the same way" (KSR; MPEP 2144.03).
3. The constant-velocity / interpolation limitation is expressly motivated by the sensor rates. R1's LiDAR (10 Hz), cameras (10 Hz) and GPS/IMU (<100 Hz) run at different rates; R3 teaches exactly how to bridge that rate mismatch — interpolate pose within a sweep under a constant angular/linear-velocity assumption and adjust every point to the sweep start time. A PHOSITA seeking to register a 10 Hz sweep to a camera frame has a strong, articulated reason to adopt R3's interpolation.
4. Common field of use and shared architectural assumptions. All of R1–R4 concern the same physical rig (rotating multi-beam LiDAR + camera + GPS/IMU on a road vehicle), the same calibration mathematics (R2/R5), and the same output (a metric 3D point cloud registered to imagery). The combination is a mere juxtaposition of known elements operating as expected — In re Keller, 642 F.2d 413 (CCPA 1981) — with no change in their respective principles of operation.

Specification admissions strengthen the case. The '101's own Detailed Description concedes that its calibration step is performed "by a method similar to that described in" R4; that its distortion correction follows R3 ("Reference in this regard can be made to 'LOAM…'"); that its camera-to-LiDAR transformation is "calibrated using a method similar to that described in" R2; and that intrinsic camera calibration follows R5. These are applicant/prior-art admissions that reduce the remaining distance between the claims and the art to the combination itself — which R1 supplies.

5. Combination B — R8 + R1 (+ R3)

Statement. R8 (DJI) teaches combining temporally sequenced laser frames into a "target reference system" using transformation rules, interpolating relative positions "to compute relative position information … at the time of any specified timestamp," re-projecting each frame's points "to the target point in time" (R8 claim 77: the target time "corresponds to an initial point of the time period" — i.e., the sweep start, exactly the '101's "starting time of the sweep"), and projecting the resulting 3D feature points into the vision sensor's images to optimise a pose-dependent target function. R1 then supplies the missing purpose element: using that LiDAR-derived correspondence across an image pair as sparse ground-truth flow.

Motivation. R8's stated problem is calibration accuracy when LiDAR and camera run at different rates and when only a subset of points is reliable (depth-discontinuous feature points, threshold-filtered). A PHOSITA building a ground-truth dataset (R1's stated goal) would naturally borrow R8's transform-to-target-time + threshold-filtered-point-selection machinery to raise ground-truth density while suppressing unreliable points. R3 supplies the residual-threshold correspondence test. Combination B is materially the same as Combination A, with R8 substituting for R3/R4 on the pose-interpolation and reference-frame transformations.

6. Dependent claims 2–13 and 14–20

7. Errors to avoid / flagged caveats

1. The "Cited By" and "Families Citing this family" tables on the Google Patents page are forward citations and are NOT prior art. US 11,587,258 B1 (GM Cruise, 2021 priority) and US 2023/0314568 A1 (Beijing Jingdong Qianshi, 2020 priority) post-date the '101 and cannot be used under §103. The same is true of the numerous CN/KR/EP family citations (2019–2022 priorities). Only WO 2018/195999 A1 (R8) qualifies as prior art from that list.
2. WO 2019/023892 A1 has a 2017-07-31 priority — post-dating the '101's 2017-06-13 priority. Excluded.
3. The seven same-day sibling US applications are not §102(a)(2) art and are commonly-owned; excluded.
4. The Unified Patents portal dates (priority 2017-06-12; "grant" 2018-12-12) remain erroneous and were not adopted; the analysis uses 2017-06-13 / 2018-12-13 (publication) / 2020-12-15 (issue).
5. Claim-drafting irregularity carried forward from the earlier section: claim 1 recites "a LiDAR scan resulting in a pair of images," and claim 14 is a method claim reciting "wherein the at least one processor is further configured to." Under the literal claim-1 reading, the "pair of images" is LiDAR-produced imagery (range/remittance images), which R1 and R4 disclose directly and which makes claim 1 more vulnerable (a §102 question). Under the intended reading (camera images), the §103 combination in §4 above governs.

8. Counterarguments the patent owner would likely raise, and their weight

• "No reference teaches using LiDAR static-point extraction as the source of ground-truth optical flow correspondences between two image frames." This is the '101's best non-obviousness argument. It is substantially weakened by R1, which does precisely that for a static-scene flow benchmark (projection of calibrated, egomotion-compensated laser points into the t₀/t₁ image pair to produce sparse ground-truth flow). Expect the patent owner to argue that R1 relied on manual correspondence selection for Velodyne-to-camera calibration and on semi-automatic outlier minimisation, whereas the '101 is automatic — but §103 asks what the combined teachings would have suggested, not whether a single reference teaches the whole.
• "Different problems" (calibration vs. benchmark generation)." R8's problem is calibration; R1's is dataset generation. Under KSR, a difference in the field of endeavour is not a bar where the references share the same sensor rig, the same coordinate-transform mathematics, and the same output data structures.
• Teaching away. I found no express teaching away in R1, R3, R4 or R8 (e.g., no statement that LiDAR-derived points should not be used as flow ground truth). R3's preference for velocity-aided distortion correction is a design choice, not a disavowal.
• Objective indicia. No evidence of record supports nexus: per the earlier litigation section, no suit asserting the '101 was identified, so there is no established commercial-success/industry-praise/nexus record to weigh against the strong prima facie case. This absence is itself notable.
• §112 leverage. The claim-14 apparatus/method hybrid may be independently invalid under §112(b), which would remove one independent claim regardless of how the §103 fight is resolved.

9. Bottom line

Weakest link: dependent claim 12's "compare local image patches / filter significant differences" — conventional in the art (left-right consistency‑style occlusion rejection) and required by the evaluation methodology of R6/R7, but I could not confirm, from the sources retrieved here, a reference using that specific formulation. Recommendation: before relying on claim 12 in any invalidity contention, run a full-text search of R6, R7, and the Scharstein–Szeliski stereo-taxonomy literature for patch-based occlusion rejection, and pull the full text of R1 (CVPR 2012, §Ground Truth) and R8/WO 2018/195999 A1 to quote the projection-and-correspondence passages verbatim.