Comprehensive analysis of intelligent robot market, industry outlook

Source：Updated：2017-02-23 08:07:28

Global intelligent robots the size of the market in 2021 is expected to grow to $33.6 billion, but growth will be the most parts of Asia. Countries in recent years, have the robot as a strategic industry, some can even be a kind of comprehensive national strength, in all areas of the leading companies are actively engaged and development.
 
 
Robots will undoubtedly has become a star of tomorrow science and technology, all around the world actively promote the industrial robot, the recent development technology, such as artificial intelligence and deep learning heat temperature, more important kinetic energy the impetus of the development of intelligent robots. According to the work being IEK study estimates, the global intelligent robots the size of the market in 2021 is expected to grow to $33.6 billion, but growth will be the most parts of Asia.
 
According to the definition of precision machinery research and development center of the financial group legal people, intelligent robot can be through the sensors to the environment, and borrowed by programmed to achieve intelligent understanding, finally reflect the required action, to carry out various production activities, services, or interact with people. It is a combination of various technologies in the platform, including mechanical, control automation, electronics, electrical, video, optical, communications, software and security system and so on related technologies and applications, including hardware and software integration technology is important. This course discusses intelligent robot industry prospect, and analyze the key technology, components and hardware and software architecture.
 
Service robot has the development potential
 
According to the institute to MIC study data display (figure 1), 2015, the big four applications robots the size of the market total about $26.9 billion, with $11 billion of industrial robot, the highest proportion, but in the whole market will be expanded to $2025 in 66.9 billion, despite the size of the market with the $24.4 billion industrial robots, but business with robots and individuals with robots in 2000-2025 by compound growth rate (CAGR) of 11.6% and 17.4% respectively, the institute for MIC industry analysts Zhang Jiahui (figure 2), points out that especially after 2015, the two types of applications grow more significant, service-oriented application market, there are many in the past did not import the emerging field of robot, to drive its growth potential.
 
Institute for MIC industry analysts Zhang Jiahui pointed out that since 2015, a service-oriented application market, there are many in the past did not import the emerging field of robot, to drive its growth potential.
 
Japanese software bank big push into robot field in recent years, a series of actions cause attention Zhang Jiahui said, including the 2012 acquisition of the French company AldebaranRobotics humanoid robot, its humanoid robot was launched in 2014, Pepper and IBMWatson, MicrosoftAzure cooperation in succession. Softbank put forward on the basis of communication to provide family and commercial application of vision, the Pepper is set to "hope to be loved" robot, through the interactive communication understand family members, being part of a home; And on the basis of artificial intelligence, let the Pepper to assist enterprise product marketing, both in the family entertainment and learning effect. In addition a well-known, now on the market and Leka and Savioke humanoid robot.
 
Countries in recent years, have the robot as a strategic industry, the development of robots for a long time, in 2015 the Japanese government initiative will set up the robot revolution, promote the development of the industrial robot. South Korea is dominated by industry trade resources, each five-year basic plan, goal is to become a robot use in 2022 countries, the production scale of 25 million Korean won; The United States started in 2011, dominated by the national science foundation (NSF), the development can safely work together with people of robotics. With the main force to develop domestic robots currently Korea, the United States is in disaster defense leading countries, Zhang Jiahui advice, layout in the family application and public have deep and commercial application of the recent rise, Taiwan can cut.
 
Deep learning, such as speech recognition technology is a significant development in recent years, thus contributing to the rise of industry and application of humanoid robot, the robot one-way communication execute commands from the past, evolved to understandable semantic response conversation, application service for robot on further development. Application situation of multiple robots, on different occasions must be combined with professional knowledge and understanding of user requirements in various fields, so the manufacturer should be through the open platform, to speed up the application in every field of robot.
 
The great leap forward artificial neural network technology
 
Deep learning because this noun in 2016 South Korea chess ai AlphaGo streak, machine the first successful challenge to the human brain, and this is generally considered the difficulty of the highest in go game activities, and by the public. And AlphaGo deep learning core is artificial neural network technology, as early as 1943, WarrenMcCulloch and WalterPitts, first put forward the mathematical model of neurons, then in 1958, psychologist Rosenblatt puts forward the concept of Perceptron (Perceptron), joined the training in the structure of the former neurons correction parameters of mechanism, the artificial neural network is the basic theoretical framework of complete. Artificial neural networks of neurons in fact from the front to collect all kinds of signals (similar to neural dendrites), then after each signal according to the weight weighted aggregation, is converted to new signal is sent through the activation function (similar to neuronal axons).
 
Related technical architecture is in the early 1970 s has been completed, the data decision technology officer Allan yiin said (figure 3), deep learning is another way of saying that artificial neural network, the success comes from a deeper understanding of the human brain works. Convolution neural network (ConvolutionalNeuralNetwork) real visual assist machine development, one of the two principle is: the local perception and weight to be Shared. Let the machine can understand the overall meaning from debris features, and then find out the characteristics of the cluster, constant refinement of hierarchical analysis, no matter how subtle features: as long as it is turned to ash, can extract the characteristics.
 
Among them, the pattern recognition is one of the key, in the cognition of the past, the central processing unit (CPU) and graphics processor (GPU) dealing with different operation function, single for deep learning function of pattern recognition, the efficiency of the GPU is CPU hundred to one thousand times, Allan yiin said further, through the depth study, machine can even remove the original Mosaic photos reduction. However, in the voice and text on the identification of Chinese for the machine, or a big challenge, Chinese words more than millions, without protocols can create new words and give the part of speech, there are many in Britain, China and Japan, China and South Korea with vocabulary, such as: blue thin, xianggu mushroom, 94, etc.
 
Giant gathering data layout in the future
 
Robot industry prospects have been attention of all parties, especially the deep learning, artificial intelligence become the giant layout the direction of the next wave of enterprise development, including Facebook (Facebook), Microsoft (Microsoft), Google (Google) and Amazon (Amazon). These enterprises in common is all through the product, service, interacting with customers, and accumulated many years of primary data, the future of artificial intelligence and deep learning, after all is a large amount of data collecting, sorting and classification, the label (Tag) that the primary data into information, and then through the result of the powerful processor search and fast response.
 
From such architecture to observe, to realize network information managing Qiu Ren twinkle (figure 4) believe that four giant FB data between the highest degree in structured, because each user upload photos or articles, has been the content arrangement, high resolution images even annotate directly, the characters in the picture the future FB to borrow from these data for further arrangement or use, can spend the least time, or to a higher quality of finishing. Currently 80% of the world's data is unstructured data, cognitive operation to enhance and simplify the learning process.
 
Robot, therefore, to reduce the error rate is highlighted in the integrity of the data and structured, Qiu Ren twinkle, to further explain the application of deep learning course from the underlying neural network computation, a large amount of data analysis, found the rules/automatic classification, produce media/recommended strategy, records user actions and feedback to the model/improve accuracy, the final is to improve data quality and self learning correction mechanism.
 
Collaborative robot more important role
 
Industrial automation from the 80 s and 90 s, intelligent manufacturing development in the 2000 s, the future of smart factory in addition to a high degree of automation and replace more complex human process, also want to develop more collaborative robot, of electronic machine business group of the robot head Peng Zhicheng (figure 5), including some glue Dispensing, screw (phrase), welding, Soldering, Inspection (Inspection), automatic Assembly (Assembly), card box truck (Pick&PlaceVehicle, P&P), etc., and even to replace flexible fixtures and tools and the production line can due to the need of products, processes and elastic restructuring (Reconfigurable), can handle more customized orders in real time.
 
A robot system, Peng Zhicheng explanation, can be divided into simple mechanical structure (Mechanicalstructure), driver, operation and control unit, sensor, communication module several projects. Collaborative robot in the future market trends for the industry, many research institutes are bullish on its development, BarclaysEquity research pointed out that the market scale will challenge the $2020 in 3 billion, the compound growth rate of 97% in 2015-2020, is a very high growth markets.
 
Leading industrial 4.0 a autonomous robots
 
The development of the intelligent robot to shout out from Germany's industrial 4.0 slogans to observe, Taiwan university associate professor, institute of automation and control Li Minfan (figure 6), said the slogan in the spirit of a small amount of variety, short cycle, all business activities begins with the seller, the traditional value chain of production order, starting from the customer demand driven research and development, to the supply chain and production, complete the whole business. Production line, therefore, must maintain highly elastic, in response to a variety of different needs, and real-time response.
 
(AutonomousMobileRobot) is different from the autonomous mobile robot autonomous robots, Li Minfan pointed out that the independent nature such as remote control is controlled by human in the cable or wireless way; Is planning a good mission computer program automatically; And independent is in unknown environment, deal with the unpredictable, and may at any time adjust the work content, stochastic elasticity. In simple terms, industry is 4.0 3.0, combined with artificial intelligence, the behavior of autonomous robots include, obstacle avoidance, target search, trajectory tracking, and maintaining the formation.
 
Sensor application quality
 
From the perspective of critical components, department of electronic engineering at hkust Lai Wenzheng (figure 7), says Dr Robot is composed of many times system integration, if further lowered system apart, can be roughly divided into displays, input devices such as keyboard/joystick, such as motor drive, communication equipment, the sensing module, vision module, navigation module, and other important units. These modules may appear multiple times in a robot body, and a microcontroller (MCU) and the sensor, will be repeated on many subsystems or modules, it is very important to the key components, and the function, the use amount of these key components will also be more remarkable.
 
Sensors, in particular, to make more intelligent robots, must through the front sensors to collect more information, stmicroelectronics (ST) simulation, micro-electro-mechanical bilateral with sensor component application manager (figure 8) shows that the sensor into action, the environment, sound categories, hot Pepper, for example, the head has four microphones, two RGB camera, a 3 d sensor, three touch sensor, chest have a gyroscope have two touch sensors, hand, leg has two sonar sensor, laser sensor six, three bumper sensor, a gyroscope, two infrared sensors, a total of nine categories nearly 30 star sensors.
 
But also because of the demand is higher and higher, sensing environmental sensors will develop into new applications, micro-electro-mechanical types of electronic compass, accelerometer and gyroscope, microphone, etc will continue to improve sensing accuracy, also is to reduce the noise, enhance sensor sensitivity and accuracy. In addition to meet the development trend of sensor fusion (SensorFusion), will lead to a homogenous sensor fusion, such as integrated accelerometer gyroscope are six shaft sensor, accelerometer integrated electronic compass and gyroscope become nine shaft sensor. The future of the single function of the sensor will be less and less.
 
Intelligent robot test project of heavy and complicated
 
Intelligent robot especially with delayed their popularity with service robot is an important reason why security, some robots at dozens of kilograms, and movement speed, if an accident at home, with children and the old man's house, I'm afraid haven't agreed with the purpose of the first lead to disaster. In addition, in the production of factories, online every day interacting with machinery, safety risk is increasing.
 
So robot product risk and validation is an important issue, UL energy and power of science and technology business development manager Chen Limin (figure 9), points out that UL1740 standard is the basic safety design requirements for robot device, the test point of power input and maximum load current, operating temperature, overvoltage and undervoltage, leakage current and operating software, handheld, conductor failure, ventilation motor deadlock deadlock, brake motor, driving motor, the components damage, overload, power and restart, emergency stop device, emergency stop time and mobile distance measurement, under the power of emergency mobile, gripping device, the teaching mode of the low speed motion measurement, by voltage protection and insulation resistance voltage withstand voltage, circuit boards, bending and axial rotation resistance, lift and move, shell strength and so on dozens of projects.
 
And it was not only the whole machine to test, Chen Limin stressed that component security and software security is tested, and many other necessary with the necessary safety protection, and the types of products related to individual features, such as the function of security (FunctionalSafety) certification, content refers to the safety and reliability of the equipment is divided into five, durable number from ten thousand to ten million. Verify the contents of intelligent robot trival, according to industry and the development of The Times, when we interact with the robot is more and more close, an unexpected accident may happen because of the new, make project more and more security verification, so if you would like to engaged in intelligent robot field, early understanding complex security certification and import the content is very important.