Patent Landscape of FA/Robotics — Fanuc's Slowdown, Bosch's Rapid Rise, and the Volume of Chinese Universities
Series Table of Contents
Introduction: Overall Map of Perspectives
Part 1: Patent Application Trends in 7 Fields
Part 2: Where Does AI/ML Have an Impact?
Part 3: Competitive Landscape of FA/Robotics (This Article)
Part 4: Competitive Landscape of Process Control
Part 5: Tech Giants' Incursion into Manufacturing
Part 6: 15 Japanese Manufacturing Companies
Part 7: Economic Value of Patents
Part 8: Patent Map of Semiconductor Foundries
Why delve into FA/Robotics?
In Part 2, FA/Robotics was determined to be a Golden Zone—where AI/ML penetration is low (3.5%), but citation boost is high (2.2x–2.7x). Despite ML-intersecting patents accounting for 15.6% of the top 1% of citations, the overall ML ratio is only 4.7%.
This gap means that patents incorporating AI/ML into FA/Robotics are few but have exceptionally high influence. So, who holds those ML patents? Who is ahead, and who is lagging behind?
Using BigQuery data, we analyze the top 40 applicants and major companies in the FA/Robotics field over the past 10 years.
Top Applicants in FA/Robotics — Surprising Names
The top 40 companies by number of patent families in the FA/Robotics (B25J + G05B19/418) field are shown, along with their ML intersection rates.
Rank Applicant Country FA Count ML Count ML Ratio 1 Fanuc Japan 2,900* 129 4.4% 2 Gree Electric China 1,253 9 0.7% 3 LG Electronics South Korea 1,232 96 7.8% 4 Samsung Electronics South Korea 966 58 6.0% 5 Siemens Germany 966 88 9.1% 6 Seiko Epson Japan 824 6 0.7% 7 ABB Switzerland 793 28 3.5% 8 Kawasaki Heavy Japan 693 12 1.7% 9 Zhejiang University China 642 25 3.9% 10 Harbin Institute of Technology China 641 7 1.1% 11 Tsinghua University China 624 23 3.7% 12 UBTech Robotics China 604 7 1.2% 13 State Grid China 590 4 0.7% 14 Omron Japan 588 38 6.5% 15 South China University of Technology China 512 23 4.5% 16 Huazhong University of Science and Technology China 477 17 3.6% 17 Haier China 466 2 0.4% 18 Boeing USA 426 10 2.4% 19 Yanshan University China 415 0 0.0% 20 KUKA Germany/China 409 8 2.0% — Mitsubishi Electric Japan 399 30 7.5% — Bosch Germany 396 96 24.2% — Panasonic Japan 392 17 4.3% — Rockwell USA 382 35 9.2% — Hitachi Japan 365 26 7.1% — Toyota Japan 362 22 6.1% — Sony Japan 339 38 11.2% — Canon Japan 341 13 3.8%
* Fanuc is the combined total of "FANUC CORP" + "FANUC LTD". FA-related CPC applications since 2015
Four structures can be seen from this table.
Structure 1 — The sheer volume of Chinese universities
There are 8 Chinese universities in the Top 40. Zhejiang University (642), Harbin Institute of Technology (641), Tsinghua University (624), South China University of Technology (512), Huazhong University of Science and Technology (477), Yanshan University (415), Shandong University (387), and Shanghai Jiao Tong University (383).
The total for these 8 universities is 4,481. This is 1.5 times that of Fanuc alone (2,900).
Even more noteworthy are the Chinese corporate groups such as Gree (1,253), UBTech (604), State Grid (590), Haier (466), and Midea/KUKA (409 + Midea 394). China is waging a total war involving both universities and companies in the FA/robotics field.
However, the ML ratio is generally low. Gree 0.7%, State Grid 0.7%, Haier 0.4%, Yanshan University 0.0%. While the volume is high, integration with AI/ML has not progressed. A "conventional-led" application pattern is prominent.
Structure 2 — A company's true competitiveness as seen through the ML ratio
The ranking of FA counts and the ranking of ML ratios paint completely different pictures.
ML比率でソート:
Bosch 24.2% ████████████████████████
Sony 11.2% ███████████
Rockwell 9.2% █████████
Siemens 9.1% █████████
LG 7.8% ████████
Mitsubishi E 7.5% ████████
Hitachi 7.1% ███████
Omron 6.5% ███████
Samsung 6.0% ██████
Toyota 6.1% ██████
Fanuc 4.4% ████
Panasonic 4.3% ████
Canon 3.8% ████
ABB 3.5% ████
Boeing 2.4% ██
KUKA 2.0% ██
Kawasaki 1.7% ██
Seiko Epson 0.7% █
Bosch's 24.2% is an outlier. One in four of the company's FA patents intersects with AI/ML.It can be read as a declaration of being an "AI/ML company" in the manufacturing FA sector.
Sony's 11.2% is also unique. While it ranks in the middle with 339 FA patent applications, it holds the second-highest ML ratio. The intersection of image sensor technology and AI/image recognition is clearly manifesting in the FA field as well.
Japan's "Big 4 FA traditionalists" (Fanuc, Yaskawa, Kawasaki, Seiko Epson) all have ML ratios of 5% or less. On the other hand, Omron (6.5%), Mitsubishi Electric (7.5%), and Hitachi (7.1%) are companies that possess IT/AI technologies outside of FA, and that difference is reflected in their patents.
Structure 3 — Fanuc's Deceleration, Bosch's Rapid Rise
Annual trends reveal the most shocking structural changes.
Fanuc — FA applications are decreasing, and the ML ratio is also declining
年 FA件数 ML件数 ML比率
──────────────────────────────
2015 204 3 1.5%
2016 319 12 3.8%
2017 429 28 6.5% ← ピーク
2018 439 27 6.2%
2019 446 23 5.2%
2020 259 17 6.6%
2021 356 12 3.4%
2022 331 10 3.0%
2023 283 5 1.8%
2024 145 1 0.7%
Fanuc's FA applications peaked at 446 in 2019 and fell to 145 in 2024. The number of ML applications dropped from 28 in 2017 to 1 in 2024.
Immediately after recording an ML ratio of 6.5% in 2017, the ratio began to fall. Fanuc stepped into AI/ML once, but it appears they have since withdrawn. Is this a management decision to deprioritize ML, or a change in filing strategy (shifting from patents to know-how)? In any case, the trajectory of Fanuc as told by patent data shows a typical "withdrawal pattern" of a sharp decline after a peak.
Bosch — From zero to 43%
年 FA件数 ML件数 ML比率
──────────────────────────────
2015 22 0 0.0%
2016 36 0 0.0%
2017 37 3 8.1%
2018 46 3 6.5%
2019 81 22 27.2%
2020 94 38 40.4% ← ピーク
2021 92 40 43.5%
2022 112 41 36.6%
2023 109 25 22.9%
2024 48 7 14.6%
Bosch's FA x ML, which was zero in 2017, reached 43.5% in 2021. In four years, they converted over 40% of their FA applications to AI/ML. This is a level of concentration unseen in other companies, converting over 40% of FA applications to AI/ML.
Bosch's "FA entry" itself has been in full swing since 2019 (46 applications in 2018 → 81 in 2019), and they have been designing FA patents with AI/ML as a prerequisite from the start. This is a fundamentally different approach from existing FA companies that "bolt on" ML.
ABB and KUKA — A contrasting decade
ABB KUKA
年 FA件数 ML比率 FA件数 ML比率
─────────────────────────────────────────
2015 82 1.2% 158 0.0%
2017 80 2.5% 129 0.0%
2019 135 6.7% 91 1.1%
2021 173 6.4% 78 3.9%
2023 144 3.5% 76 2.6%
ABB doubled its FA applications from 82 to 173, and its ML ratio also temporarily rose to 6.7%. They are advancing AI integration while growing.
KUKA is the opposite. It halved from 158 to 76 applications. After being acquired by Midea Group in 2016, the pace of applications has clearly dropped. Although the ML ratio is rising (0% → 2.6%), it is negligible in terms of absolute numbers.
Structure 4 — What is the content of FA x AI/ML?
We break down the specific technologies where AI/ML is applied in the FA field using G06N sub-classifications.
2015 2016 2017 2018 2019 2020 2021 2022 2023 2024
────────────────────────────────────────────────────────────────────
深層学習 42 94 203 293 488 585 540 567 550 632
汎用ML 27 55 126 191 342 407 404 387 328 259
知識ベース 20 31 58 91 141 129 134 114 88 123
確率モデル 12 22 30 43 89 74 63 57 46 39
Deep learning (G06N3: Neural Networks) is the largest and only growth area, with 632 applications in 2024. General-purpose ML (G06N20) peaked in 2020 and is declining. Knowledge-based (G06N5) and probabilistic models (G06N7) are following the same trend.
The patent battlefield for FA x AI is becoming consolidated into deep learning. Image recognition (visual inspection, picking), reinforcement learning (robot motion planning), and anomaly detection—in all these cases, deep learning-based technologies are replacing conventional methods.
Composition of technical sub-domains by company — "What" are they patenting?
Looking at the CPC sub-classifications for FA/robots by company reveals the strategic position of each firm.
Company B25J9 Control G05B19/418 FA Control B25J13 Sensing B25J15 Gripper B25J19 Accessory Fanuc 1,335 332 510 196 575 Siemens 360 868 51 31 62 KUKA 469 22 134 96 194 ABB 484 259 98 66 157 Omron 297 250 129 42 86 Bosch 235 129 48 38 40 Mitsubishi Electric 146 231 63 21 41 Yaskawa 281 80 83 — 97
Two completely different approaches are visible.
Robot Body Type (Fanuc, KUKA, Yaskawa): B25J9 (Robot program control) is the largest. They are patenting the robot's movement itself—trajectory planning, speed control, and collision avoidance.
FA Control Type (Siemens, Mitsubishi Electric, Omron): G05B19/418 (Factory adaptive control) is the largest. They are patenting factory-wide control logic—MES integration, production scheduling, and quality feedback—rather than the robot itself.
Siemens' G05B19/418 = 868 patents is the highest among all companies. The company's FA strategy is not to "make robots" but to "control the entire factory including robots." This is the intellectual property manifestation of their Totally Integrated Automation (TIA) / Xcelerator strategy.
ABB is the only company with a foot in both camps. With high levels in both B25J9 (484 patents) and G05B19/418 (259 patents), they cover both the robot body and factory control.
Collaborative Robots (cobots) — Volume has grown, but AI has not yet entered
年 cobot件数 ML交差 ML比率
──────────────────────────────────
2015 214 1 0.5%
2017 480 2 0.4%
2019 619 8 1.3%
2021 685 13 1.9%
2023 738 13 1.8%
Collaborative robot-related patents have grown 3.4 times from 214 to 738. However, the ML intersection rate is only 1.8%. This is even lower than the 3.5% for FA as a whole.
Safety (force limiting compliant with ISO/TS 15066) and physical design (lightweighting, flexible joints) are at the center of patents, while AI/ML remains at a "nice to have" level. The patent race for cobots is still in the hardware phase.
Summary — 5 facts revealed by the FA/robot patent map
1. Fanuc leads in number of applications but is retreating in AI/ML After recording an ML ratio of 6.5% in 2017, it fell to 0.7% in 2024. Total applications have also halved. There is a risk of entering a "withdrawal pattern" of sharp decline after a peak.
2. Bosch has become the overwhelming leader in FA × AI/ML Their ML ratio of 24.2% is the highest among manufacturing FA companies. By using a "digital-native FA" approach that designs FA patents with AI as a prerequisite from the start, they have created a different playing field from traditional FA companies.
3. China dominates the "volume" of FA through an all-out effort by universities and companies The total for 8 universities + home appliance/infrastructure companies is more than double that of Fanuc. However, the ML ratio is generally low, with the majority being "conventional FA" applications.
4. Siemens and Fanuc are fighting in completely different domains Siemens = Factory control (G05B19/418), Fanuc = Robot body (B25J9). The "what" of their FA patents is completely different. Only ABB has a foot in both.
5. FA × Deep Learning is the only growing AI field General-purpose ML, knowledge bases, and probabilistic models are all in decline. Only deep learning continues to grow in 2024. The competition in FA × AI will be decided by the ability to apply deep learning.
Data source: Google Patents Public Dataset (BigQuery: patents-public-data.patents.publications). Covers applications from 2015 onwards. CPC classification: B25J (Manipulators) + G05B19/418 (Factory adaptive control). AI/ML intersection: Co-occurrence with G06N (Machine Learning).
This series is written and published in the flow of Markdown → HTML → note. Google Patents Public Dataset (BigQuery) was used to obtain patent data, and Claude Code was used for analysis and writing.
This article represents personal views and does not represent the views of any specific company or organization.
