Humanoid Robot Whole-Body Motion Planning Patent Landscape 2026 | Patsnap
- 2025 was the breakout year. A single year accounts for the majority of filings in this corpus — confirming that serious commercial-scale investment in whole-body motion planning is very recent, not a settled field.
- US and China dominate receiving offices. Nearly all families land in either the United States or China, with South Korea a distant third — meaning filings outside those two jurisdictions leave wide open space for regional portfolio positioning.
- The most-cited record is not a trajectory paper. The single highest-cited document concerns character animation matching, not control theory — a signal that cross-domain influence from entertainment and graphics is shaping how practitioners think about motion realism.
What whole-body motion planning means for humanoid robotics IP
Whole-body motion planning coordinates every joint of a humanoid or legged robot simultaneously — balancing centre-of-mass trajectory, foot placement, and upper-body posture under real-time constraints. Unlike arm-only manipulation or fixed-base control, it must satisfy dynamic feasibility, terrain geometry, and task objectives all at once. Patent claims in this space therefore tend to be layered: a footstep planner claim sits on top of a 3-D environment model, which in turn depends on a depth-sensing or SLAM front-end. That layering makes freedom-to-operate analysis non-trivial, because blocking a middle layer can obstruct a whole downstream stack.
The IPC composition of this corpus reflects that layering clearly. The majority of records carry B25J (manipulators and robots) as the primary class, but a significant share also carry B62D — the vehicle-steering class that covers bipedal locomotion kinematics — alongside G05D (non-electric variable control) and G06N (AI-based computing). A record tagged with all four sits at the intersection of mechanical design, control law, and learned policy, which is precisely where the hardest-to-design-around claims live. Publication lags filing by roughly 18 months, so 2025's peak figure understates true activity; families filed in late 2025 will not be fully visible until 2027.
A field that accelerated sharply and is still mid-climb
Twenty patent families across a decade of search coverage sounds modest, but the distribution is almost entirely back-loaded. The corpus was near-dormant through the early part of the period and compressed most of its output into the last two years — a pattern that signals a field in formation rather than one approaching saturation. For R&D teams deciding where to file, that back-loading is an opportunity: claim space that would be crowded in adjacent robotics sub-fields is still relatively open here.
Annual filing trend: from dormant to peak
Filing activity was essentially zero through 2017 and remained thin through 2021. Growth inflected around 2022, and 2025 registered the highest single-year count in the corpus at ten families. The most recent year is partial, so its figure of one should not be read as a decline; the 18-month publication lag means families filed in 2025 and 2026 are still emerging from examination. The honest read is that the field's momentum remains upward, not that it has reversed.
Technology composition: robots first, AI close behind
B25J dominates with 19 of 20 records, confirming that all major players anchor their claims in the robot-hardware class before layering on control or learning sub-classes. B62D's presence in seven records is notable — it reflects claims that treat bipedal locomotion as a vehicle-dynamics problem, which opens prosecution in a less-crowded examination art unit. G06N's three records represent the reinforcement-learning and neural-network claims that are attracting the most recent filer attention, and they are the sub-area most likely to grow rapidly as learned controllers move from lab to product.
Go deeper on Humanoid Robot – Whole-Body Motion Planning with Eureka
This page is one run against one query. Ask Eureka your own question about humanoid robot – whole-body motion planning and every answer comes back with the patent numbers behind it.
Try EurekaThe records that structure prior art in this space
Footstep Planning Method, Robot and Computer-Readable Storage Medium
The method acquires depth images of the walking environment, constructs a three-dimensional model from those images, identifies planar candidate regions within the model, and selects among them to generate footstep locations. The claim chain ties perception directly to placement decision, meaning a design-around must either avoid depth-based environment modelling or re-route the candidate-selection logic outside the claimed pathway.Patent number, assignee and grant/publication date are shown in the data table above. Abstract is editorially paraphrased for landscape context; refer to the primary document for legal claim language.
| # | Patent | Citations |
|---|---|---|
| 1 | Method for developing and controlling a robot to have movements matching an animation character | 31 |
| 2 | Footstep Planning Method for Bipedal Robot | 18 |
| 3 | 基于在线质心轨迹优化的人形机器人高动态跳跃运动控制方法 | 10 |
| 4 | Learning robust legged robot locomotion with implicit terrain imagination via deep reinforcement learning | 8 |
| 5 | Bipedal action model for humanoid robot | 6 |
| 6 | 一种人形机器人全身协调行走的运动在线生成与控制方法、系统、计算机可读存储介质及计算机程序产品 | 5 |
| 7 | Footstep planning method, robot and computer-readable storage medium | 3 |
| 8 | Planning and control method for legged robot, apparatus, robot, and storage medium | 2 |
| 9 | Bipedal action model for humanoid robot | 2 |
| 10 | 考虑滑移转向的轮腿式机器人全身运动控制系统及方法 | 2 |
Citation counts within a searched corpus favour older records because they have had more time to accumulate references. Treat these rankings as influence signals, not as a measure of technical quality or current commercial relevance.
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 filing patterns tell practitioners
With only twenty families spread across a decade, this is an early-stage corpus — but the structure of citations and IPC co-classifications already reveals where influence concentrates and where the field has unfinished business.
Animation-derived motion leads on citations
The most-cited record in the corpus concerns matching robot motion to animation characters — a cross-domain technique that roboticists have found generalisable to natural whole-body motion generation. Its citation lead over the second record (18 citations) is large enough to treat it as a genuine prior-art anchor, not an outlier.
US is the primary battleground; Korea is a watching brief
More than half of all families designate the United States as the primary receiving office, reflecting where enforcement risk is highest and where the large commercial humanoid programmes are headquartered. China's six families suggest its assignees file domestically first and internationally selectively. Korea's two families are sparse enough that a well-crafted filing programme there would face thin prior art.
Learned locomotion is the fastest-moving frontier
Only three families carry the G06N AI-computing class, yet reinforcement-learning-based locomotion is the approach generating the most published research in 2024–2025. That gap between research volume and patent volume is a classic signal of an area where claim space is still open and early filers can set broad positions.
Academic-industrial collaboration is limited but present
The only identified co-assignee pair links a Harbin Institute of Technology affiliate with its Suzhou research arm — a university-spinout structure common in Chinese deep-tech. The rarity of cross-organisation co-filing across the whole corpus suggests that most players guard their motion-planning IP closely rather than sharing development through joint ventures or consortia.
Eureka can read the same corpus for gaps instead of for coverage: under-claimed branches adjacent to humanoid robot – whole-body motion planning, with the prior art for and against each one.
| Assignee | Co-assignee | Shared families |
|---|---|---|
| HIT (Harbin Institute of Technology) Suzhou Research Institute | Harbin Institute of Technology (HIT) | 1 |
哈工大苏州研究院 and 哈尔滨工业大学 appear together on one family — the sole instance of cross-organisation co-assignment in this corpus. Every other assignee files independently, making this pair a notable exception in an otherwise fragmented ownership landscape.
Who is building positions in whole-body motion planning
Ownership in this corpus is distributed across academic institutions, national research labs, industrial robotics companies, and one major entertainment conglomerate. No single assignee holds a dominant block; the landscape is competitive at the top and trails off into a long tail of single-family entrants. That structure means freedom-to-operate analysis must cover multiple independent owners, not a single licensor.
Universities anchor the prior-art base
Several leading engineering universities — including institutions associated with Beijing, Harbin, and Yale — appear as assignees. University patents are often licensed non-exclusively and prosecuted with broad independent claims, making them more likely to appear as blocking prior art than as competitive moats held by a single commercial actor.
Commercial players prioritise perception-to-planning pipelines
UBTECH's featured record illustrates the commercial approach: claims that lock together depth sensing, environment modelling, and footstep selection into a single patentable pipeline. Disney's presence — historically in character animation — reinforces the insight that entertainment-derived motion realism is crossing into robotics IP, particularly for whole-body expressiveness.
KAIST holds a focused but thin position
South Korea's two receiving-office records align with KAIST's known humanoid robotics programme. The position is narrow relative to the institution's research output, suggesting that much of the underlying work has not been converted to granted patent families — or that filings are still pending. Either way, the Korean prior-art landscape for this topic is thin enough to allow differentiated entries.
| Assignee | Recent year | YoY |
|---|---|---|
| Digital Intelligence Company (Taiwanese entity; no established single English registration identified — conservative transliteration) | 0 | -100% |
| Korea Advanced Institute of Science and Technology (KAIST) | 0 | — |
| Disney Enterprises, Inc. | 0 | — |
| Beijing Institute of Technology (BIT) | 0 | -100% |
| Suzhou Cancon Intelligent Technology Co., Ltd. | 0 | -100% |
| Yale University | 0 | — |
| Tsinghua University | 0 | -100% |
| UBTECH Robotics Corp Ltd | 0 | — |
Where this landscape is heading
The corpus is young and the filing rate is accelerating. Practitioners tracking this space should expect significant new prior art to emerge over the next 18–24 months as the 2025 filing peak becomes fully visible and commercial humanoid programmes convert R&D into prosecution dockets.
Watch the G06N sub-class closely
Reinforcement-learning locomotion controllers are the dominant research paradigm but remain sparsely patented. As lab-validated policies move toward product deployment, companies will begin prosecuting learned-policy claims aggressively. Monitoring new G06N + B25J co-classifications is the fastest way to catch that wave early.
Set up a Eureka alert for this IPC combination →Korea and Europe are thin jurisdictions
With only two Korean receiving-office records and no European families visible in this corpus, both jurisdictions represent relatively open terrain for first-mover filings. Companies with commercial operations in those markets should evaluate whether their existing US or CN family claims translate into viable national-phase entries.
Analyse jurisdiction gaps in Patsnap Eureka →Monitor the 2025 cohort as it publishes
Ten families filed in 2025 will become fully searchable through 2026 and into 2027 as they clear the 18-month publication lag. That cohort will likely reshape the citation network and reveal which technical routes the most active filers are betting on. A landscape refresh in mid-2027 will capture it completely.
Schedule a landscape refresh in Patsnap Eureka →Practitioner questions about this landscape
By raw family count — twenty families over a decade — this is a sparse corpus compared with adjacent robotics sub-fields such as manipulation grasping or SLAM. However, sparsity does not mean openness: the families that do exist often carry broad independent claims that cover core pipeline steps like environment modelling, trajectory generation, and footstep selection. A practitioner should conduct claim-level freedom-to-operate analysis rather than relying on headline family counts to assess crowding. The most exposed areas are depth-image-based footstep planning (anchored by the UBTECH family) and character-animation-derived whole-body motion, where citation density signals that prior art is well-established.
The primary class is B25J (manipulators and robots), which covers nearly all families in this corpus. B62D/57 (bipedal locomotion treated as a vehicle-dynamics problem) is the second most important and is often missed by searchers who restrict themselves to the robot subclasses. For learned controllers, G06N (computing based on AI/neural networks) and G05D (control of non-electric variables) are the relevant supplementary classes. A complete search combines all four, because a single family may carry co-classifications spanning all of them. Restricting to B25J alone will miss roughly a third of the relevant prior art in this particular corpus.
A peak of ten families in 2025 means a substantial block of new prior art will not become fully searchable until 2026–2027, owing to the standard 18-month publication lag between filing and publication. Freedom-to-operate opinions written today therefore carry higher uncertainty than is typical, because they cannot see the claims that are currently pending from that cohort. Practitioners should flag this gap explicitly in any FTO opinion and schedule a refresh once the 2025 cohort clears publication. In the interim, monitoring pre-grant publication feeds for B25J + G06N co-classifications is the best available early-warning approach.
The United States is the dominant receiving office in this corpus, accounting for twelve of the twenty families, and the UBTECH family (US20220040859A1) is a published application with a perception-to-placement claim chain that is broad enough to warrant close reading before any product launch involving depth-image-based footstep planning. Whether any given US record has issued as a granted patent, and in what claim form, must be verified against the USPTO register — publication status alone does not confirm enforceability. The cross-domain records deriving from animation character matching also designate the US and should be reviewed for granted claim scope, particularly for products that prioritise motion naturalness or expressiveness.
Several universities appear as assignees in this corpus, including institutions associated with Beijing, Harbin, and Yale. Academic patents in robotics are frequently licensed non-exclusively through technology-transfer offices, and broad independent claims from university prosecution are a common source of licensing revenue. Before approaching any institution, it is worth mapping which specific families they own, whether those families have granted claims in the jurisdictions relevant to your product, and whether the institution has an active licensing programme or tends to monetise through sponsored research agreements instead. The Harbin Institute of Technology / Suzhou Research Institute co-assignment pair is worth particular attention because the institutional structure may mean licensing decisions are split across two entities.
The evidence points to several under-claimed branches: reinforcement-learning-based whole-body locomotion policies (only three G06N-tagged families exist), multi-contact balancing under external disturbance, and real-time footstep re-planning on unstructured terrain. Loco-manipulation — coordinating leg locomotion with arm manipulation simultaneously — appears in research literature but is essentially absent from the claim landscape in this corpus. Whole-body model-predictive control combined with learned dynamics models is another area where research activity significantly outpaces patent filings. Any of these branches offers the prospect of setting broad, defensible independent claims against a thin prior-art base, provided applications are filed before the 2025 cohort fully publishes and potentially closes some of that space.
Research Humanoid Robot – Whole-Body Motion Planning in depth with Eureka
Go past this page: query the whole humanoid robot – whole-body motion planning 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.