Humanoid Robot Collaborative Robot Patents: Leaders & White Space 2026
- Filings surged to 18 in 2025 after years of flat activity, a late and sharp step-change rather than steady growth from 2017 onward.
- B25J manipulator claims cover every record in the set, while adjacent classes like G06N and G05B appear in only a handful of families each.
- The most-cited records are recent 2026 US filings on bipedal action models, signalling where examiners and competitors are already paying attention.
What this patent set actually covers
This landscape isolates patent families at the intersection of humanoid or bipedal robot hardware and human-robot collaboration behaviours — collision detection, safe interaction and collaborative operation — narrowed to the core manipulator and control IPC classes (B25J9/16, B25J13, B25J19/06). It is a small, sharply defined set: 24 published families against a search string that excludes generic industrial-arm safety filings and generic bipedal locomotion work that never touches collaboration claims.
Because publication typically lags filing by around 18 months, the 2025 and 2026 counts in the trend chart are the least complete and will likely revise upward as later filings publish.
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 24-family set: when the filings happened, and which IPC subclasses they touch beyond the core manipulator classes that define the search.
A flat decade, then a spike
Filings sat near zero through the late 2010s and stayed low through the 2022 midpoint (2 families), before jumping to a peak of 18 in 2025. That shape points to a recent wave of entrants rather than a maturing field with steady incremental filing.
Concentration in B25J, thin elsewhere
Every family in the set touches B25J (manipulators and robots), by construction of the search. Beyond that core, B62D (motor vehicles and steering) appears in 5 families, G06N (AI-based computing) in 4, and G05B (control and regulating systems) in only 2 — with E04F, G05D, G06F and G06T each appearing in a single family.
Shares are the percentage of the 24 records in scope. A patent can carry several IPC classes, so the shares add up to more than 100%.
Go deeper on Humanoid Robot Collaborative Robot with Eureka
This page is one run against one query. Ask Eureka your own question about humanoid robot collaborative robot and every answer comes back with the patent numbers behind it.
Try EurekaThe records other filings are built on
Training and use of a bipedal action model for humanoid robot
The filing describes a hierarchical bipedal action model (BAM) for controlling a humanoid robot: a base controller trained via reinforcement learning in simulation, a Gamma model that generates intermediate goals, a Beta model that translates those goals into task-space actions, and an Alpha model that translates task-space actions and robot state into motor commands. The initial BAM is deployed with the Alpha model running on-board the robot, which then performs an initial task while logging sensor and control data.Filed by Figure AI Inc., published 2026-05-07.
| # | Publication no. | Patent title | Citations |
|---|---|---|---|
| 1 | US20260070221A1 | Bipedal action model for humanoid robot | 6 |
| 2 | US20260097492A1 | Annotation model for humanoid robot data | 5 |
| 3 | WO2025221916A1 | Advanced array of sensor assemblies of a humanoid robot | 5 |
| 4 | US20260084309A1 | System and method for calibration of humanoid robots | 4 |
| 5 | CN119973996A | 一种基于多模态感知的人形机器人智能装修施工系统及方法 | 4 |
| 6 | US20230133934A1 | Method for controlling legged robot, robot and computer-readable storage medium | 3 |
| 7 | US20260061611A1 | Management of multiple modes for humanoid robot | 1 |
| 8 | US20260124746A1 | Training and use of a bipedal action model for humanoid robot | 1 |
| 9 | US20260126804A1 | Bipedal action model for humanoid robot | 1 |
| 10 | US20260084314A1 | System and method for training and using a bipedal spatial perception model | 1 |
Citation counts here reflect influence inside this searched corpus, not overall importance — older or more foundational-sounding filings tend to accumulate citations simply by being available longer.
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 a filing decision
Four read-outs from the trend, the citation table and the technology split that matter more than the raw counts on their own.
The wave is recent and still building
Filing activity was essentially flat through the late 2010s and only reached 2 families by 2022. The jump to 18 in 2025 is the whole story of this landscape — most of what exists here was filed in the last two years, and 2026's single record so far is an artefact of publication lag, not a slowdown.
B25J is fully occupied, adjacent classes are not
Every record in the set carries a B25J manipulator classification, which is expected given the search scope. What is notable is how thin the adjacent classes are: G06N (AI models) reaches only 4 families and G05B (control systems) just 2, despite both being natural pairings for collaborative humanoid control.
Attention is already on 2026 bipedal-model filings
The most-cited record in the set is a 2026 filing on a bipedal action model for humanoid robots, followed closely by related sensor-array and calibration filings. That the newest filings are already the most-cited suggests active, current competitive attention rather than legacy citation accumulation.
Filing is concentrated in the US, with China a distant second
The United States receives more than half of the tracked filings (14 of 24), with China next at 4. India, New Zealand and WIPO/PCT each account for only 1-2 filings, indicating this is currently a US-centred filing pattern rather than a globally distributed one.
Eureka can read the same corpus for gaps instead of for coverage: under-claimed branches adjacent to humanoid robot collaborative robot, with the prior art for and against each one.
Who is filing, and how fast they are moving
Momentum in this set is uneven: several assignees that filed in prior years show no activity in the latest tracked year, while at least one entrant is active right now.
Still filing in the most recent year
Shenzhen Weixia Robot is the one assignee in this set with a filing recorded in the latest tracked year, making it the clearest signal of ongoing activity as the rest of the field shows a drop-off.
Several established names have gone quiet
UBTech, Digital Dream Labs, and a handful of other assignees each show a -100% year-over-year change, meaning they filed previously but have no record in the latest year. This does not necessarily mean they have exited the space — it may reflect filings still working through publication lag.
Co-assignment is rare in this set
Only one co-assignee pairing appears across the 24 families, and it occurs once. This is a landscape of largely independent filers rather than one shaped by joint ventures or shared-ownership filing strategies.
| Assignee | Recent year | YoY |
|---|---|---|
| Shenzhen Weixia Robot Co., Ltd. | 1 | — |
| Digital Dream Labs | 0 | -100% |
| UBTech Robotics Corp Ltd | 0 | -100% |
| SoftBank Robotics Europe | 0 | — |
| Golden Crown Intelligent Technology (Suzhou) Co., Ltd. | 0 | -100% |
| Hangzhou Fanjia Technology Co., Ltd. | 0 | -100% |
| Wuxi Zhidongli Robot Co., Ltd. | 0 | -100% |
| Symbiosis International University, India | 0 | -100% |
Where to take this analysis
The numbers above describe what has already published. Turning them into a filing or freedom-to-operate decision means going deeper on specific claims and specific assignees.
Map claim scope against your own architecture
Run your control-stack design against the claim language of the most-cited bipedal action model filings to see exactly where overlap risk sits before you draft.
Explore claim charting in EurekaTrack the assignees still active in 2025-2026
Set alerts on the small group of assignees showing live filing activity rather than the larger group that has gone quiet, since that is where near-term competitive moves will surface first.
Set up assignee tracking in EurekaProbe the under-claimed branches before they fill in
The thin adjacent classes identified here — sensor fusion, onboard calibration, control-AI integration — are exactly the areas where a well-drafted first claim still has room.
Run a white-space search in EurekaCommon questions about this landscape
This landscape tracks 24 published patent families that combine humanoid or bipedal robot terms with human-robot collaboration language, filtered to core manipulator and control IPC classes such as B25J9/16, B25J13 and B25J19/06. It is a narrow, precisely scoped set rather than a broad humanoid robotics count, since it excludes generic industrial-arm safety patents and pure locomotion filings that never address collaboration. Because publication lags filing by roughly 18 months, the true number of filed-but-unpublished families for 2025-2026 is almost certainly higher than what is currently visible.
Filing activity in this set is spread across a mix of established robotics firms and newer entrants, with no single assignee dominating the 24-family total. Momentum data shows most tracked assignees, including established names like UBTech, had no filings in the latest year, while at least one newer filer, Shenzhen Weixia Robot, shows activity right now. Co-filing between assignees is rare, with only one co-assignee pairing appearing across the entire set, suggesting this remains a field of largely independent filers rather than joint ventures.
Filing was essentially flat from 2017 through the early 2020s, reaching only 2 families by the 2022 midpoint. The real inflection came later: filings peaked at 18 in 2025, meaning the large majority of this landscape's activity is extremely recent. The single 2026 record so far understates that year's actual activity, since publication lag means many 2025-2026 filings have not yet appeared in the public record.
Every family in this set carries a B25J manipulator classification by definition of the search scope, but coverage extends into related areas at much lower density: motor vehicle and steering systems (B62D) appear in 5 families, AI-based computing (G06N) in 4, and control and regulating systems (G05B) in just 2. Single-family appearances in building finishing (E04F), non-electric control (G05D), digital data processing (G06F) and image processing (G06T) point to niche applications, such as construction-site robotics, that are only beginning to intersect with collaborative humanoid claims.
The clearest gaps sit in the classes adjacent to the core manipulator claims rather than within B25J itself, which is fully occupied across all 24 families. Multimodal sensor fusion for collision detection, reinforcement-learned task-space action translation, and onboard calibration routines for bipedal systems all appear in only a handful of families despite being natural technical companions to the core collaboration claims. A first claim drafted around integrating an AI-based motion model directly with a control-and-regulating subsystem, an area with only 2 families in this set, would currently face limited direct prior art within this landscape.
Research Humanoid Robot Collaborative Robot in depth with Eureka
Go past this page: query the whole humanoid robot collaborative robot 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.