Humanoid Robot Machine Vision Patents: Who Leads, Gaps 2026
- Filing only just started. the trend goes from zero in 2017 to 24 families in 2025, with growth still accelerating through the 2022 midpoint — this is an early-innings field, not a mature one.
- B25J dominates the IPC mix. 35 of 39 families sit in manipulators & robots, versus single digits for G06T image processing and G06V recognition — vision claims are still framed as robot subsystems, not standalone perception.
- Momentum is already turning over. several assignees that filed in prior years show 0 in the latest year or -100% YoY, while only a couple of names still show fresh activity — leadership here has not settled.
A young, narrow filing base with vision claims nested inside robot-body IPCs
Humanoid robot machine vision, as defined by this search, covers 39 patent families published between 2015 and mid-2026, filed against a search string that combines humanoid/bipedal/legged robot terms with robot vision, 3D perception, object recognition and visual servoing. The filing trend is close to flat until the early 2020s, then climbs sharply to 24 families in 2025, the peak year of the dataset. That shape says the underlying technology — vision systems built specifically for bipedal or legged robot bodies, rather than generic machine vision — has only recently become a distinct filing target.
The IPC composition reinforces this reading. B25J (manipulators and robots) appears in 35 of 39 records, far ahead of G06N (AI models, 10), G06T (image processing, 9) and G06V (recognition, 5). Most applicants are still filing vision claims as part of a robot-body patent rather than as a freestanding perception patent, and receiving-office data shows the United States and China together account for the bulk of filings, with India, WIPO, Europe and Japan trailing well behind.
Filing trend and technology composition
Two views of the same 39-family dataset: how filing volume has moved year over year, and how those filings distribute across IPC subclasses.
Filings accelerate from a standing start
Zero families in 2017, one at the 2022 midpoint, then a jump to 24 in 2025 — the most recent year (2026) is partial and will understate true filing volume once publication catches up, given the roughly 18-month lag between filing and publication.
Vision claims still ride inside robot-body filings
B25J leads with 35 records, well ahead of G06N (10), G06T (9), B62D (6), G06V (5), G01C (4) and G05D (4) — a distribution that shows perception technology is mostly claimed as a feature of the robot rather than as an independent vision system.
Shares are the percentage of the 39 records in scope. A patent can carry several IPC classes, so the shares add up to more than 100%.
Go deeper on Humanoid Robot Machine Vision with Eureka
This page is one run against one query. Ask Eureka your own question about humanoid robot machine vision and every answer comes back with the patent numbers behind it.
Try EurekaMost-cited families and a representative filing
System and method for training and using a bipedal spatial perception model
A humanoid robot system pairs vision sensors with a bipedal spatial perception model: a feature extractor builds hierarchical, multi-scale feature maps via a feature pyramid network, a robot data module detects robot parts within those maps, and a robot vector data module calculates 3D spatial position and orientation for each detected part by predicting 2D-to-3D point correspondences and solving perspective geometry.Filed by Figure AI Inc., published 2026-03-26.
| # | Publication no. | Patent title | Citations |
|---|---|---|---|
| 1 | CN120516701A | 一种人形机器人多模态指令解析系统 | 10 |
| 2 | CN118628802A | 多模态特征融合图像分类方法及在人形机器人中的应用 | 7 |
| 3 | US20260070221A1 | Bipedal action model for humanoid robot | 6 |
| 4 | US20260097492A1 | Annotation model for humanoid robot data | 5 |
| 5 | WO2025221916A1 | Advanced array of sensor assemblies of a humanoid robot | 5 |
| 6 | JP2009042145A | Apparatus and method for object recognition | 5 |
| 7 | US20260084309A1 | System and method for calibration of humanoid robots | 4 |
| 8 | US20250353195A1 | Balance control method and apparatus for wheel-legged robot, device, and storage medium | 4 |
| 9 | CN119927906A | 一种用于人形机器人上肢协同控制的交互方法及装置 | 4 |
| 10 | CN117841002A | 一种基于多模态模型的人形机器人实时抓取路径规划方法 | 4 |
Citation counts reflect influence within the searched corpus and skew toward older records; recent high-value filings may not yet have accumulated citations.
Patent titles are shown in the language they were filed in, not translated, so that each record stays verifiable against the original filing — a translated title will not match in Eureka or in any national register. 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 numbers mean for filing strategy
Three read-outs from the trend, IPC and citation data that matter more than the raw counts.
The field is still in its accelerating phase
With the 2022 midpoint at a single family and 2025 at 24, growth has not plateaued. Filing this early in a curve means claim scope is still negotiable in most sub-areas — the dense zones are B25J-adjacent, not vision-specific.
Vision is claimed as a robot feature, not a standalone system
G06T (9) and G06V (5) trail B25J by a wide margin. Applicants targeting perception-only claims — independent of a specific robot mechanism — face less occupied space than the headline B25J number suggests.
Filing activity concentrates in two offices
The United States and China together account for the large majority of receiving-office activity, with India, WIPO, EPO and Japan each in single digits. Freedom-to-operate work should prioritise US and Chinese prior art first.
Last year's filers are not this year's filers
Several assignees active in prior years show zero filings in the latest year, some down -100% YoY, while only a small number show continued activity. No single assignee has established durable year-over-year momentum yet.
Eureka can read the same corpus for gaps instead of for coverage: under-claimed branches adjacent to humanoid robot machine vision, with the prior art for and against each one.
Who is filing, and where the gaps sit
The assignee base is small and its recent-year momentum is unsettled — several names that filed in prior years have gone quiet, and no assignee shows sustained year-over-year growth into the latest period.
Small counts, large swings
Assignees with any latest-year activity are filing at a rate of one family, against prior-year bases that make the percentage changes look dramatic (-93% YoY in one case) despite the small absolute numbers.
Multiple prior filers show zero in the latest year
A cluster of assignees that filed previously — spanning both Chinese research institutions and specialist robotics firms — recorded no filings in the most recent year, consistent with a field where individual entrants file once or twice rather than sustaining programmes.
No dominant assignee yet
At 39 total families with no single name capturing a large share, this looks like an open field: a research-heavy tail of universities and specialist firms rather than a small number of entrenched incumbents.
| Assignee | Recent year | YoY |
|---|---|---|
| Digital Artificial Intelligence Corporation | 1 | -93% |
| Zhejiang University | 1 | — |
| Tencent Technology (Shenzhen) Co., Ltd. | 0 | -100% |
| Jiangsu Yunmu Zhizao Technology Co., Ltd. | 0 | — |
| Kuawei (Shenzhen) Intelligent Digital Technology Co., Ltd. | 0 | -100% |
| Jianghuai Frontier Technology Collaborative Innovation Center | 0 | -100% |
| Li Guanghui | 0 | — |
| NSK Ltd. | 0 | — |
Where to take this analysis
The dataset points to an early-stage, geographically concentrated field with claim space still open outside the core B25J cluster.
Run a freedom-to-operate check on the most-cited families
Start with the highest-cited US and Chinese records before drafting new claims in bipedal spatial perception or multi-modal instruction parsing.
Explore in Patsnap EurekaTrack assignee momentum quarter by quarter
Several filers have already dropped to zero year-over-year; watching who re-enters will show where competitive investment is actually landing.
Set up monitoring in Patsnap EurekaCommon questions about humanoid robot machine vision patents
This dataset contains 39 patent families published between 2015 and mid-2026 that match a search combining humanoid, bipedal and legged robot terms with robot vision, 3D perception, object recognition and visual servoing. Filing volume was near zero until the early 2020s and reached 24 families in 2025, the peak year recorded so far. Because publication typically lags filing by around 18 months, the 2026 figure is still partial and will rise as more applications publish.
The assignee base is fragmented, with no single company holding a dominant share of the 39 families. Recent-year momentum data shows several assignees, including Chinese research institutions and specialist robotics firms, filing once and then dropping to zero in the following year rather than sustaining a filing programme. This pattern is typical of an early-stage field where corporate and academic labs are still testing claim positions rather than defending established territory.
The IPC distribution shows B25J (manipulators and robots) in 35 of the 39 families, making it by far the dominant classification, followed by G06N (AI models, 10 records), G06T (image data processing, 9) and G06V (image and video recognition, 5). Smaller counts appear in B62D (motor vehicles and steering, 6), G01C (navigation and gyroscopes, 4) and G05D (control of non-electric variables, 4). This means most filings frame vision technology as part of a robot mechanism rather than as a standalone perception system, which leaves room for claims that isolate the perception layer.
The gap between the 35 B25J records and the much smaller G06T (9) and G06V (5) counts suggests vision-specific claims — independent of a particular robot mechanism — are less crowded than the headline numbers imply. Sub-areas such as 2D-to-3D point correspondence for robot-part pose estimation, multi-modal instruction parsing, and sensor-array configurations for full-body 3D perception show thinner filing density relative to the core manipulator cluster. A first, carefully scoped claim in one of these branches faces less occupied prior art than a claim filed squarely inside B25J.
US20260084314A1, filed by Figure AI Inc. and published 2026-03-26, is a representative recent filing: it describes a bipedal spatial perception model using a feature pyramid network to build multi-scale feature maps, a module that detects robot parts within those maps, and a vector module that calculates each part's 3D position and orientation by solving 2D-to-3D point correspondence. It illustrates the trend toward robot-specific perception architectures rather than generic computer-vision techniques adapted after the fact. Anyone drafting claims in bipedal pose estimation should review its scope closely before finalising claim language.
Research Humanoid Robot Machine Vision in depth with Eureka
Go past this page: query the whole humanoid robot machine vision 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.