In view of the lack of patent big data in research on technology foresight in the industrial robot field, this paper introduces an improved method based on patent mining and knowledge map. Firstly, SAO structure is extracted from selected patents, secondly, the similarity between patents is calculated based on extracted SAO structure, thirdly, patent network and patent map are drawn based on calculated patent similarity matrix, technology evolution process and future trends of industrial robot are summarized from patent network, and future potential technology opportunities are predicted based on technological vacancies identified from patent map. Finally, this paper identifies six key technical areas and four potential technical opportunities in the field of the industrial robot.
In the 21st century, when the role of scientific and technological progress in the engine of economic development is becoming more and more obvious,science and technology competition between countries is intensifying, how to filter many technical fields and concentrate resources on areas of greatest potential for development,it’s a matter of great concern to both enterprises and governments, technology foresight is an effective way to solve this problem. In addition, the inevitable nature of the application of technology makes it important to study the impact of new technologies in advance, so it is necessary to carry out technology foresight activities.
China has clearly defined its strategic transformation goal toward “Strong Manufacturing Country”. At present, the target is still relatively broad and vague, and there is no in-depth technical analysis, especially the technology foresight of industrial robot. The manufacturing industry of China urgently needs to promote intelligent manufacturing and achieve upgrading. Through technology foresight, the development direction of industrial robot technology will be condensed; the industrial robot industry will be guided to achieve industrial innovation and technological progress, and key technologies that will lead the future to be identified for strategic focus and R&D investment, which is important for improving the R&D capability of smart equipment in China.
Technology foresight is a systematic process that identifies future technological developments and their interactions with the economy, society, as well as environment, and provides actionable guidance for creating a better future [
At present, technology foresight methods can be summarized as qualitative analysis based method, quantitative analysis based method, and multi-method combination method. Although quantitative analysis based method of technology foresight has developed a lot, the current method is still dominated by qualitative analysis and supplemented by quantitative analysis, with strong subjectivity, so more scientific and effective quantitative analysis based methods need to be developed to make up for the deficiency of qualitative analysis based methods. Brockhoff pointed out that patents contain 90% to 95% of all the world’s latest scientific and technological knowledge [
Patent analysis refers to the processing of some patent information in patent literature by means of data processing or statistical analysis, so as to provide organizations with valuable information that can comprehensively understand and predict a certain field [
The statistical analysis based on digital attributes mainly analyzes the external digital features of patent documents, such as statistical analysis of patent number for applicants, application years, and applicant countries. Chen Xin et al. [
A common shortcoming of statistical analysis based on numerical attributes is the loss of high-value unstructured patent document text information. Therefore, more and more researches begin to focus on the method of patent text mining to deeply explore the content of patent texts [
The International Organization for Standardization (ISO) defines industrial robot as three-axis or multi-axis multi-purpose manipulator oriented at automatic control and reprogrammable in the field of industrial automation [
According to the above definition, industrial robot refers to mechanical devices that can replace human labor through various automated operations in the industrial field, and is usually used in handling, welding, spraying and assembly, etc [
At present, the technical level and popularization scale of industrial robots in China are quite different from those of developed countries represented by Japan, the United States, Germany, and South Korea. Taking the density of robot as an example,
By reading these key technical literatures, the related patents were retrieved with industrial robot as the key word. Since it usually takes nearly two years to review a patent application before it is authorized, most of the patents will be rejected during this period. In order to improve the quality of the patent data, this paper
only retrieves the authorized patents, and the patent retrieval type compiled is as follows: @(title,abstract, claims, body) ((“industr* robot*”~1) or (“industr* manipulat*”~1)) @(title,abstract, claims) (“robot*”or”manipulat*”) @* (kind_code_b or kind_code_b1 or kind_code_b2 or kind_code_c or kind_code_y or kind_code_u) (@* inno_utility_patent), where “@(title, abstract, claims, body)” and “@(title, abstract, claims)” represent defined search scope, “((“industr* robot*”~1) or (“industr* manipulat*”~1))” and “(“robot*” or “manipulat*”)” represent keywords to be searched within the specified search scope, and “(“industr* robot*”~1)” represents fuzzy matching, and the rest of the suffixes are the screening of patent types.
By using the retrieval method above, a total of 18,100 pieces of authorized patent data were obtained on the patent analysis platform, and the data were downloaded to the local database as the basic data for subsequent analysis.
By analyzing the application time of 18100 patent data, the number of patent applications in each year is shown in
1) Concept presentation stage (1954-1987). In the 1930s, the concept of robot began to appear. And in 1954, American George Dvor applied for the world’s
first industrial robot patent “a programmed article transfer”, providing a programmable joint transfer material device. This stage belongs to the germination period of technology and is also the popularization period of industrial robot related concepts, and the market scale is small. In the late 1970s, many companies from Europe, America, and Japan began the global deployment of industrial robot patents.
2) Extinction stage of upsurge (1987-2005). At this stage, the number of patents showed a trend of decline year by year. Main reasons for the decline were the lack of unified industry standards, the failure to break the barriers of cross-industry cooperation, the patent monopoly of important production technologies, and the dislocation between product value and public demand.
3) Rapid development stage (2006-present). This stage is marked and driven by breakthroughs in artificial intelligence, machine learning and other related technologies, as well as increased policy support for the development of industrial robot industry in various countries. For example, in the “Industrial Transformation and Upgrading Plan (2011-2015)”, the Chinese government has identified the development direction of intelligent manufacturing equipment, including accelerating the development of industrial robots such as welding, handling, and assembly robot.
The patentee here refers to the entity that currently enjoys the patent right, which can be citizens, collective ownership units, foreign trade enterprises, and sino-foreign joint ventures. A total of 5482 patents are identified from 18,100 patent data, as shown in
3370 patents are held by schools and specialized research institutions, and 2583 patents are held by individuals, the patent holders of the remaining 288 patents are missing. The corporate sector holds more than half of the patents related to industrial robot, indicating the active commercialization of industrial robot.
Among all the patentees, the top 20 patent holders are listed in
In this paper, the patent strength algorithm of the Innography patent analysis platform is used to select patents with strength greater than 2, so the low-value patents are excluded. Finally, 7247 patent data are obtained and stored in a local database.
SAO structure can be automatically extracted from the text using existing natural language processing tools. Currently, there are three softwares for extracting SAO structure, including Knowledgist, AlchemyAPI and Stanford Parser. Stanford Parser is a syntactic parser developed by Stanford University that can recognize English, Chinese and Arabic and is a fully open source natural language processing tool. In this paper, Stanford Parser is adopted as SAO extraction tool, and the SAO structure is extracted by the software package offered by Java development environment IntelliJ IDEA. Before extracting SAO structure, original patent data should be cleaned firstly. Finally, 65,225 SAO are
Ranking | Organization | Ranking | Organization |
---|---|---|---|
1 | Fanuc Ltd. | 11 | Denso Corporation |
2 | ABB Ltd | 12 | Seiko Epson Corporation |
3 | Kuka AG | 13 | Toshiba Corporation |
4 | YASKAWA Electric Corporation | 14 | Siemens AG |
5 | Mitsubishi Electric Corporation | 15 | Fraunhofer |
6 | Hitachi, Ltd. | 16 | Honda Motor Co., Ltd. |
7 | Panasonic Corporation | 17 | Kobe Steel, Ltd. |
8 | Samsung Electronics Co., Ltd. | 18 | Chinese Academy Of Sciences |
9 | Kawasaki Heavy Industries, Ltd. | 19 | Toyota Motor Corporation |
10 | Sony Corporation | 20 | Canon Inc. |
extracted from 7247 patents, and finally a total of 62,738 SAO structures are obtained after cleaning.
This paper uses the Lin algorithm in JAVA’s JWS open source project, combined with WordNet semantic dictionary, to calculate the similarity between vocabulary. Lin algorithm is the most widely used classical semantic similarity algorithm recently. The lexical similarity calculation is shown in Equation (4-1):
L i n ( W 1 , W 2 ) = 2 ∗ d e p t h ( l c s ( W 1 , W 2 ) ) d e p t h ( W 1 ) + d e p t h ( W 2 ) (4-1)
where, depth(Wi) is the distance from the vocabulary Wi node to its root concept node, lcs(W1, W2) represents the minimum common subset of vocabulary W1 and W2, and Lin(W1, W2) has the minimum value of 0, indicating that the two terms are completely different, And the maximum value of 1, indicating that the two terms are exactly the same.
Based on the similarity calculating method between words, Lin algorithm is adapted to calculate the similarity between two sentences, as shown in Equation (4-2):
S i m ′ ( X , Y ) = 2 ∗ M a t c h ( X , Y ) | X | + | Y | (4-2)
where, S i m ′ ( X , Y ) represents the similarity of two sentences, and its value is 0 to 1. The larger the value is, the higher the similarity between two sentences will be. Match(X, Y) represents the logarithm of the same feature in sentence X and sentence Y, and 2 times represents the number of the same feature; | X | and | Y | represent the number of features in sentence X and sentence Y respectively.
SAO has a complete sentence structure, subject S and object O are mostly phrases composed of multiple words, representing products, technologies or technical indicators. Therefore, the similarity of each part of the SAO structure should be calculated separately, and then the comprehensive similarity can be obtained, as shown in Equation (4-3):
S i m ′ ( S A O i , S A O j ) = [ S i m ′ ( S i , S j ) + S i m ′ ( A i , A j ) + S i m ′ ( O i , O j ) ] (4-3)
where, S i m ′ ( S i , S j ) , S i m ′ ( A i , A j ) and S i m ′ ( O i , O j ) respectively refer to the similarity of S, A and O combination in SAOi and SAOj. When the value reaches a certain threshold value t, it is deemed that SAOi and SAOj are completely the same; otherwise, they are considered different. The judgment formula is shown in Formula (4-4):
S O i j = { 1 , S i m ( S O i , S O j ) ≥ t 0 , other (4-4)
Then, the similarity between patent A and patent B can be calculated, as shown in Equations (4)-(5):
S i m A , B = 2 ∗ M a t c h ( A , B ) | A | + | B | (4-5)
where, Sim(A, B) is the patent similarity, the higher the value is, the higher the similarity is. When the value is 1, it means that the two patents are identical. | A | and | B | represent the number of SAO in patent A and B respectively, and Match(A, B) represents the logarithm of the same SAO in patent A and B.
According to the above method, the similarity between patents is calculated based on the extracted SAO structure, then the patent similarity matrix is obtained, as shown in
This paper uses the Ucinet analysis tool to draw a patent network for the industrial robot patent data. In order to facilitate the identification of the core patents, the display threshold t of the edge is set here, and the edge below the threshold t will not be shown in the patent network, besides, isolated points will not appear in the network also.
From
P1 | P2 | P3 | P4 | P5 | P6 | P7 | P8 | P9 | P10 | P11 | P12 | P13 | P14 | P15 | |
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
P1 | 0.000 | 0.493 | 0.404 | 0.400 | 0.368 | 0.458 | 0.391 | 0.333 | 0.396 | 0.430 | 0.435 | 0.369 | 0.395 | 0.409 | 0.482 |
P2 | 0.000 | 0.532 | 0.412 | 0.398 | 0.485 | 0.440 | 0.446 | 0.511 | 0.483 | 0.483 | 0.452 | 0.339 | 0.554 | 0.481 | |
P3 | 0.000 | 0.424 | 0.446 | 0.469 | 0.372 | 0.352 | 0.502 | 0.450 | 0.391 | 0.452 | 0.459 | 0.516 | 0.505 | ||
P4 | 0.000 | 0.393 | 0.383 | 0.367 | 0.369 | 0.391 | 0.378 | 0.407 | 0.406 | 0.357 | 0.405 | 0.414 | |||
P5 | 0.000 | 0.407 | 0.373 | 0.359 | 0.431 | 0.388 | 0.387 | 0.446 | 0.275 | 0.451 | 0.451 | ||||
P6 | 0.000 | 0.400 | 0.357 | 0.522 | 0.452 | 0.423 | 0.428 | 0.226 | 0.490 | 0.396 | |||||
P7 | 0.000 | 0.402 | 0.400 | 0.392 | 0.348 | 0.370 | 0.361 | 0.399 | 0.404 | ||||||
P8 | 0.000 | 0.329 | 0.389 | 0.305 | 0.285 | 0.316 | 0.335 | 0.361 | |||||||
P9 | 0.000 | 0.359 | 0.415 | 0.410 | 0.384 | 0.382 | 0.501 | ||||||||
P10 | 0.000 | 0.435 | 0.403 | 0.401 | 0.466 | 0.457 | |||||||||
P11 | 0.000 | 0.344 | 0.397 | 0.431 | 0.395 | ||||||||||
P12 | 0.000 | 0.404 | 0.479 | 0.423 | |||||||||||
P13 | 0.000 | 0.426 | 0.307 | ||||||||||||
P14 | 0.000 | 0.476 | |||||||||||||
P15 | 0.000 |
According to
In terms of sensing technology, the types of sensors are constantly enriched, and the sensing systems of robots are becoming more and more complex. From the initial force, position and speed sensors to the current proximity sensing, machine vision and other intelligent sensing technologies, industrial robot of awareness are becoming more and more close to human limits even greater than that of human perception.
In terms of network communication, early wired communication technologies gradually developed wireless communication technologies including WIFI, Zigbee, and Bluetooth suitable for different occasions. From wireless communication to wired communication, the efficiency of information transmission is higher and the energy consumption is lower, and the scope of communication has been greatly expanded.
In terms of control technology, the early stage of program control is mainly based on teaching reproduction control technology. With the continuous development of sensing technology, the adaptive control method with feedback function can be realized, making industrial robots have stronger environmental adaptability; therefore, the application scenario is broadened. At present, intelligent control represented by fuzzy control, optimal control, neural network control is the current research hot spot.
In terms of programming technology, the fixed program generated by the
Core Patent Number | Key SAO | Technology Theme |
---|---|---|
P1816 P4976 P1264 P1364 P335 P24 | [Position sensor-produce-output signal]; [Vision unit-photograph-guide line]; [Safety sensor-comprise-contact-free proximity sensor]; [Pair of camera-obtain-camera image coordinate system]; [Industrial grinding robot-have-visual sense and touch sense]; [Proximity sensor-detect-safety-relevant part of working chamber]; [Pair of camera-obtain-camera image coordinate system]; [Sensor model simulate sensor imaging]. | Sensing Technology |
P639 P4264 P849 P862 P1306 | [Master controller and intelligent terminal-be in wireless communication connection with-cloud server]; [Receiver-receive-wireless information]; [System comprise ZIGBEE local area network module]; [Robot body include Bluetooth module information store unit]; [Automatic charge grind equipment comprise RFID label array]. | Network Communication Technology |
P826 P962 P591 P1639 P3548 P641 P1705 P123 | [Controller-recognize-position by image]; [Control device-monitor-state of photosensor]; [Cloud control mode improve communication efficiency]; [Fuzzy controller-control-automatic guiding vehicle]; [Invention-disclose-trajectory tracking control method]; [System-comprise-numerical control subsystem]; [Position and speed data on teaching point-be supply to-arithmetic controller]; [Invention-disclose-multifunctional screw tighten system and control method]; [Regular patterning code locating controller-control-tray label reading code]. | The Control Technology |
P409 P2066 P1721 P2025 P1500 P1394 | [Position and speed data on teaching point-be supply to-arithmetic controller]; [Teaching-produce-new control information]; [Path programming technology-design-cleaning path]; [Invention provide complex path programming method]; [Invention discloses offline programming and modifying method]. | Programming Technology |
P5011 P962 P706 P3217 P4531 | [Handheld teaching unit comprise interface]; [Operation adopt human-computer interface]; [AGV trolley is connected with PC through PLC and wireless WiFi]; [Graphic editing interface is used for interactive graphic programming between user and system]. | Interactive Technology |
P1257 P5010 P911 P3976 | [Intelligent moving and cleaning robot-be used in -large scale cleaning equipment]; [Automatic guided system-include-automatic guided vehicle]; [Invention-provide-mobile robot]; [Welding main body-slide on-guide rail device]. | Mobile Technology |
traditional online teaching programming technology has poor adaptability to the actual situation and needs to occupy the working time of the robot, so it is gradually replaced by offline programming and semi-automatic programming technology. Semi-automated programming can reduce manual intervention and increase the automation and intelligence level, and will move toward self-programming in the future, which is also the key technology to realize intelligent technology.
In terms of interactive technology, from the early teaching box to the current intelligent terminal, the interactive devices are more and more diverse, and the way of interaction is more and more humanized. In the future development, human-computer interaction will further develop in a user-friendly direction, such as further development of intelligent interaction technology that controls robots through gestures, language and expressions.
In terms of mobile technology, from the beginning of the rail movement to the current industrial drones, the mobility of industrial robots is getting stronger and stronger, the range of movement is getting larger and larger, and the number of movable paths is increasing. With the development of mobile technology, the energy consumption of devices and the security between devices will become more and more prominent.
The patent map of industrial robot is shown in
Vacancy 1: Intelligent Sensing System. Most of the patents around Vacancy 1 involve sensing and measurement-related technologies to improve the accuracy of data acquisition and achieve sensor integration. Robots can’t be isolated; it must be integrated with the environment to work properly. The idea of IoT is to create an organism that interconnects machines, environment and people together.
Vacancy 2: Generalization, Standardization, Networking. The development trend of the future robot technology is generalization, standardization and networking. On the one hand, it can facilitate the robot information interaction between different manufacturers, realizes long-distance operation monitoring, maintenance and remote control. On the other hand, it can also reduce the industry cost. Modular and flexible robot has strong adaptability to the environment and task, and its cost advantage is obvious, which will lead the development of high-tech manufacturing in the future.
Vacancy 3: Reducing energy consumption and developing green technology. The patents around vacancy 3 are mainly related to the technology of power supply for industrial robot system. At present, wireless power supply (P489-JP6076426, Automatic guided vehicle and power supply system) has emerged. If the technology can further reduce the drive and servo motor energy consumption, it is bound to bring new development opportunities. In addition, it can also be combined with solar energy and other new clean energy to power equipment, in line with the environmental protection-oriented technology development ideas in 21st century.
Vacancy 4: Remote Management and Control. The technical subject described in patents around vacancy 4 is mainly related to remote monitoring, remote diagnosis, and remote control operation. Remote management and control can achieve physical isolation between workers and industrial work site and provide higher security. In addition, the development of remote control technology has made it possible to conduct industrial activities in all kinds of harsh environments, such as deep-sea mining operations, mineral exploitation and so on.
A number of representative industrial robot companies have emerged in China, but their output value is low and cannot compare with the industrial robot multinational companies represented by the “four major families” in terms of technology, scale and R&D investment. Therefore, an effective way for domestic companies to compete with multinational companies is to adopt a patent strategic alliance to compete with multinational companies with the power of groups and alliances.
It is an important measure to make full use of big data analysis technology and timely grasp the development trend of industrial robot technology in the world to avoid the patent trap of industrial robot power. Through the analysis of big data from the three aspects of patent, market and law, the author gives an early warning on the threat situation of industrial robot patent, the development trend of key parts and technologies of industrial robot, the dynamics of main competitors and possible patent infringement disputes.
The robot application in the automobile industry has been occupied by foreign robot giants, which is related to the early development of the automobile industry in developed countries. In China, new industries such as high-speed railway and new energy are the most competitive fields for Chinese robots. China has accelerated the application of automation in 3C, ceramics, household appliances, logistics and other emerging fields, bringing more opportunities to robot enterprises. A comprehensive opportunity is to develop specialized industrial robots for the needs of niche segmentation.
Industrial robots have become the strategic commanding heights of the world in high-tech competition and are listed as a priority technology. Although China has included industrial robots in the strategic emerging industry catalogue and one of the 100 major projects and projects planned to be implemented during the 13th Five-Year Plan period in China, compared with Japan, Korea, and other countries, the industrial policy support system is still incomplete. Government should increase support and improve the industrial robot industry development policy.
Based on patent mining and knowledge mapping methods, this paper conducts a technology foresight study on the technological development in the field of industrial robot. In the aspect of text information extraction, the natural language processing open source library is comprehensively used, and the patent similarity calculation method based on SAO is proposed. Finally, with the help of patent network and patent map visualization tools, six key technical fields and four potential technical opportunities for industrial robot are summarized, and suggestions for the development of industrial robot in China are put forward finally.
The authors declare no conflicts of interest regarding the publication of this paper.
Wen, X.H. (2019) Technology Foresight Research of Industrial Robot Based on Patent Analysis. Journal of Data Analysis and Information Processing, 7, 74-90. https://doi.org/10.4236/jdaip.2019.72005