ML State of Health Estimation Patents: Who Leads, Where Gaps Are 2026
- 34,448 records in scope, with filings peaking at 1,994 in 2025 before the most recent, still-incomplete year.
- No dominant gatekeeper: the leader holds 509 records and the top 10 combined account for just 6.4% of all records in scope.
- Diagnosis and imaging classes overlap heavily with core AI computing, with A61B, G06V, G06K and G06T all sitting alongside G06N and G06F in the same records.
What this landscape covers
This landscape covers 34,448 published records filed or published between 2015 and mid-2026 that combine state-of-health language, capacity fade estimation or battery health prediction with claim-level detail on feature extraction, duty cycles, on-board computation, memory constraints, data labelling or uncertainty quantification. The search string deliberately pairs a subject-matter term with implementation-level claim language, so the corpus skews toward records that specify how a state estimate is produced rather than merely referencing battery state in passing.
Because publication lags filing by roughly 18 months, the 2026 count of 510 understates actual filing activity for that year; treat the most recent one to two years as a floor, not a ceiling.
Let an AI agent run this analysis on your own technology
Pick a task. Every answer cites the patents behind it.
Filing trend and technology composition
Two views of the same 34,448 records: how filing volume has moved year over year, and which IPC subclasses the records fall into.
Filings rose then plateaued
Annual filings moved from 890 in 2017 to a peak of 1,994 in 2025, passing through a 2022 midpoint of 1,442. That trajectory is flat-to-declining rather than accelerating once the incomplete 2026 count of 510 is set aside as a lag artefact rather than a real drop.
Diagnosis and imaging classes sit alongside core AI computing
G06F (13.5% of records) and G06N (11.1%) anchor the corpus as expected for a machine-learning topic, but A61B diagnosis and surgery claims appear in 7.5% of records, and image/data-recognition classes G06V, G06K and G06T each clear at least 4% — evidence that health-estimation claims are frequently drafted alongside sensor and imaging pipelines rather than as standalone algorithms.
Shares are the percentage of the 34,448 records in scope. A patent can carry several IPC classes, so the shares add up to more than 100%.
Go deeper on Machine Learning State of Health Estimation with Eureka
This page is one run against one query. Ask Eureka your own question about machine learning state of health estimation and every answer comes back with the patent numbers behind it.
Try EurekaA representative record
US20220217214A1 — Feature extraction device and state estimation system
A feature extraction device that extracts a feature quantity for input into a state estimation model, built from an acquisition unit that collects time-series activity logs, an extraction unit that derives co-occurrence relationships between activities within a specified period, and a generation unit that aggregates activity feature quantities.Filed by NTT DOCOMO, published 2022-07-07.
| # | Publication no. | Patent title | Citations |
|---|---|---|---|
| 1 | US6752816B2 | Powered surgical handpiece with removable control switch | 1,918 |
| 2 | US6705503B1 | Electrical motor driven nail gun | 1,572 |
| 3 | US20060178918A1 | Technology sharing during demand and supply planning in a network-based supply chain environment | 1,303 |
| 4 | US6090123A | Powered surgical handpiece with state marker for indicating the run/load state of the handpiece coupling asse… | 1,284 |
| 5 | US20040105264A1 | Multiple Light-Source Illuminating System | 1,261 |
| 6 | US6366813B1 | Apparatus and method for closed-loop intracranical stimulation for optimal control of neurological disease | 1,117 |
| 7 | US9368991B2 | Distributed battery power electronics architecture and control | 1,027 |
| 8 | US5404960A | Motor-driven power steering apparatus for automobiles | 1,022 |
| 9 | US6280381B1 | Intelligent system for noninvasive blood analyte prediction | 953 |
| 10 | US20160001781A1 | System and method for responding to driver state | 926 |
High citation counts here reflect age and general-purpose applicability more than direct relevance to state-of-health estimation specifically; several of the most-cited records predate the topic's current framing and cover adjacent surgical or supply-chain subject matter picked up by the search terms.
Each row carries its publication number; clicking a row searches Eureka by that number.
Put your own technology through the same analysis
Eureka on the web
When you want the answer in the next five minutes.
The agent works the prompt against patents and technical literature, citing every source.
Run your analysis now →MCP server & REST API
When it has to run inside your own pipeline.
Patent search, landscape analysis and assignee resolution as MCP tools. Drop them into any agent framework, or call REST directly.
Browse MCP servers →What the data shows
Three patterns stand out once volume, concentration and classification are read together.
No single assignee controls the field
The leader holds 509 records and the fifth-place assignee holds 187 — a steep drop-off, but even the combined top 10 account for only 6.4% of all records in scope. That leaves the overwhelming majority of filings distributed across a long tail of single- and few-filing entrants.
Recent-year activity is cooling among the largest filers
Several of the most active assignees show sharp year-on-year declines in their latest-year counts, while at least one shows a small increase from a low base. Given publication lag, this cooling should be read cautiously, but the direction is consistent across multiple large filers rather than isolated to one.
Claims cluster around general computing and AI-model classes
G06F and G06N together anchor the largest share of records, but A61B diagnosis claims (7.5%) and three separate imaging/recognition classes each exceeding 4% show that health-estimation claims are routinely drafted with sensor or imaging context rather than as pure algorithmic claims.
Filing is US-heavy with meaningful China and EPO volume
The United States receives the largest single share of filings at 8,963, with China at 3,548 and the EPO at 2,977. WIPO/PCT filings of 2,066 indicate a meaningful share of applicants pursuing multi-jurisdiction protection from the outset.
Eureka can read the same corpus for gaps instead of for coverage: under-claimed branches adjacent to machine learning state of health estimation, with the prior art for and against each one.
Who is filing, and where the gaps sit
The ranked leaders span consumer electronics, telecom infrastructure and diversified technology conglomerates rather than automotive OEMs or battery cell makers specifically — a sign that much of the claim activity approaches state-of-health estimation as a general machine-learning or sensing problem rather than a battery-domain one.
A clear leader without a controlling share
The top-ranked assignee holds 509 records, well ahead of fifth place at 187, but this leadership does not translate into control of the field given the top 10's combined 6.4% share of all 34,448 records.
Co-filing is limited and concentrated
Only 10 co-assignee pairs appear in the data, with the strongest pairing linking two related corporate entities under a shared parent. Cross-company joint filing is not a defining feature of this field so far.
Momentum is mixed, not uniformly declining
Most of the largest recent-year filers show double-digit percentage declines, but at least one shows a large percentage increase off a small base count of 2. Momentum data should be read alongside absolute counts, not on percentage change alone.
| Assignee | Recent year | YoY |
|---|---|---|
| Huawei Technologies Co., Ltd. | 5 | -58% |
| Qualcomm Inc. | 3 | -85% |
| Samsung Electronics Co., Ltd. | 3 | -75% |
| Intel Corp. | 2 | +100% |
| Tencent Technology (Shenzhen) Co., Ltd. | 2 | -33% |
| Empire Technology IoT Portfolio 2016 LLC | 1 | — |
| International Business Machines Corporation | 0 | -100% |
| General Electric Company | 0 | -100% |
Where to take this
The dataset points to specific next steps depending on whether the goal is freedom-to-operate, competitive tracking or claim drafting.
Check freedom-to-operate on under-claimed branches
Memory-constrained on-board inference and uncertainty-quantified capacity fade estimation show lighter density than the core AI computing classes — worth a targeted search before drafting.
Run a freedom-to-operate search in EurekaTrack momentum among cooling large filers
Several leading assignees show sharp year-on-year declines; confirm whether this reflects a genuine pullback or a publication-lag artefact before drawing conclusions about competitive intent.
Set up assignee monitoring in EurekaMap claim overlap between diagnosis and computing classes
A61B, G06V, G06K and G06T claims frequently co-occur with G06N/G06F claims in the same records — a structured claim chart across these classes will show where combination claims are getting granted.
Build a claim chart in EurekaCommon questions
Among the 100 companies in the ranked assignee list, the leading assignee holds 509 records, with the fifth-place assignee at 187 and the tenth at 152. However, the top 10 combined account for only 6.4% of all 34,448 records in scope, so no single company or small group controls the field. The remaining records are spread across a long tail of assignees with far fewer filings each, which matters for freedom-to-operate work since risk is distributed rather than concentrated in one or two portfolios.
Filing volume rose from 890 in 2017 to a peak of 1,994 in 2025, passing through 1,442 at the 2022 midpoint, which is a flat-to-declining trajectory rather than sustained acceleration. The 2026 figure of 510 looks like a sharp drop but is almost certainly a publication-lag artefact, since publication typically trails filing by around 18 months. The safest reading is that filing activity plateaued around 2022-2025 rather than continuing to climb.
G06F (electric digital data processing) and G06N (AI-model computing) are the largest classes, covering 13.5% and 11.1% of the 34,448 records respectively. A61B (diagnosis and surgery) covers 7.5%, and three imaging or recognition classes — G06V, G06K and G06T — each cover more than 4% of records. Because a single record can carry several IPC classes, these shares add up to more than 100%, which is expected and reflects how often health-estimation claims combine sensing, imaging and computing elements in one filing.
The IPC composition suggests lighter claim density around memory-constrained on-board inference, uncertainty quantification applied specifically to capacity fade, duty-cycle-conditioned feature extraction, weak-label training approaches for degradation data, and cross-chemistry transferable models. These are inferred from the relative weighting of classes and filing patterns in this dataset rather than from a class-by-class gap count, so a targeted prior-art search on the specific implementation is still the necessary next step before drafting.
US20220217214A1, filed by NTT DOCOMO and published 2022-07-07, claims a feature extraction device that builds a state estimation model from time-series activity logs, using an extraction unit that derives co-occurrence relationships between activities and a generation unit that aggregates activity feature quantities. It is drafted around a general activity/state estimation framework rather than being specific to battery capacity fade, so its blocking effect on a battery-specific state-of-health claim depends heavily on how narrowly or broadly its claim terms are construed. Anyone drafting in this space should read its independent claims in full rather than relying on the abstract.
Research Machine Learning State of Health Estimation in depth with Eureka
Go past this page: query the whole machine learning state of health estimation corpus yourself, in your own scope.
Every answer comes back with patent numbers you can open.
Disclaimer. This page is generated from Patsnap Eureka data drawn from a limited snapshot of global patent and scientific-literature records, and is provided for general information and reference only.
Patent data carries inherent limitations: recent filings (typically the most recent 18–24 months) are under-counted due to standard publication lag; counts may be reported at either a patent-family or a patent-record basis and are not always directly comparable; classification, applicant-name, and citation data may contain errors, duplicates, or omissions; and the underlying search query defines and constrains the scope shown. As a result, the analysis may be incomplete or inaccurate and may not reflect the full technology landscape.
Nothing on this page constitutes an exhaustive prior-art, novelty, freedom-to-operate, or validity search, nor does it constitute legal, financial, investment, or professional advice, and it should not be relied upon as such. Any patent, commercial, or strategic decision should be verified independently and reviewed with qualified patent, legal, and domain professionals. Patsnap makes no warranties, express or implied, as to the accuracy, completeness, or fitness for any particular purpose of the information presented.
Machine translation. Assignee and organisation names originally recorded in Chinese, Japanese or Korean have been rendered into English by an AI translation step so that the tables stay readable. These renderings are best-effort and may not match a company’s registered English name; the original name is what the underlying patent record carries, and it is what any Eureka query launched from this page uses.