103,99 €
Mehr erfahren.
- Herausgeber: John Wiley & Sons
- Kategorie: Wissenschaft und neue Technologien
- Sprache: Englisch
INTELLIGENT SATELLITE DESIGN AND IMPLEMENTATION Integrate cutting-edge technology into spacecraft design with this groundbreaking work Artificial intelligence and machine learning have revolutionized virtually every area of computing and complex engineering, and the design of satellite spacecraft is no exception. Intelligent satellites are increasingly capable of human-like perception, decision-making, and operations, and their problem-solving capacities are still expanding. As AI and machine learning continue to advance, their integration into satellite manufacture will only deepen. Intelligent Satellite Design and Implementation seeks to understand the foundations of this integration and its likely directions in the coming years. Beginning from the basic principles of interaction between artificial intelligence and satellite design and mission planning, the book analyzes a series of current or potential areas of technological advancement to create a comprehensive overview of the subject. Intelligent Satellite Design and Implementation readers will also find: * Background information on the introduction and development of artificial intelligence * Detailed discussion of topics including autonomous satellite operation, remote sensing satellites, and many more * Over 100 illustrations and tables to reinforce key concepts Intelligent Satellite Design and Implementation is ideal for graduate students and advanced undergraduates in engineering, computing, and spacecraft design programs, as well as researchers in these and related fields.
Sie lesen das E-Book in den Legimi-Apps auf:
Seitenzahl: 360
Veröffentlichungsjahr: 2023
Ähnliche
Table of Contents
Cover
Table of Contents
Series Page
Title Page
Copyright Page
About the Authors
Preface
1 Development of Artificial Intelligence
1.1 The Concept and Evolution of Artificial Intelligence
1.2 The Current Scope and Technical Framework of Artificial Intelligence
1.3 The Overall Development Trend of Artificial Intelligence
1.4 The Main Achievements of AI
References
2 Artificial Intelligence in the Satellite Field
2.1 The Concept and Connotation of Intelligent Satellite
2.2 Technical Characteristics of Intelligent Satellite System
2.3 Opportunities and Challenges for Satellite to Develop AI
References
3 Development Status of AI Technology in Satellites
3.1 Policy and Planning
3.2 Technology and Application
3.3 Development Trend Analysis
References
4 Basic Knowledge of AI Technology
4.1 The Concepts and Characteristics of Machine Learning and Deep Learning
4.2 Key Technologies of AI
4.3 Machine Learning
4.4 Natural Language Processing
4.5 Knowledge Engineering
References
5 AI Requirements for Satellite System
5.1 Demand Requirements for AI Technology in Satellite System
5.2 Challenges and Solutions of Artificial Intelligence in Aerospace Applications
References
6 Intelligent Remote‐Sensing Satellite System
6.1 Technical Analysis of Intelligent Remote‐Sensing Satellite System
6.2 Basic Structure of Intelligent Remote‐Sensing Satellite System
6.3 Key Technical Directions of Intelligent Remote‐Sensing Satellite System
6.4 Typical Application Cases
References
7 Intelligent Communication Satellite System
7.1 Requirements for AI System Technology by Communication Satellite
7.2 Key Technologies of Communication Satellite Intelligent System
7.3 Typical Application Cases
References
8 Intelligent Navigation Satellite System
8.1 Intelligent Management of Constellation Network
8.2 Satellite Independent Health Management
8.3 Intelligent Fault Diagnosis and Prediction Technology
8.4 Intelligent On‐Orbit Maintenance of Satellite
8.5 Typical Application Cases
References
9 Application of AI in Aerospace Loads
9.1 Intelligent Load Software Architecture
9.2 Cloud Service Center Software Architecture
9.3 Network‐oriented Communication Protocol
9.4 Intelligent Expert System
9.5 Intelligent Execution System
9.6 Intelligent Semantic Interpretation System
9.7 Intelligent Load Onboard Intelligent Processing Technology Scheme
9.8 Digital Multi‐function Load
References
10 Future Development of Intelligent Satellite
10.1 Application Prospect of AI in Aerospace Field
10.2 The Next Development Gocus
10.3 Summary
References
Index
End User License Agreement
List of Tables
Chapter 2
Table 2.1 Classification of intelligent reasoning/intelligent control by AI...
Chapter 3
Table 3.1 Relevant policies and regulations issued by advanced aerospace co...
Table 3.2 Summary of main satellite application intelligent technology.
Table 3.3 Three modes of human–computer cooperation.
Chapter 4
Table 4.1 Key AI technologies and their definitions.
List of Illustrations
Chapter 1
Figure 1.1 Three waves of AI.
Figure 1.2 Key points of AI technology development.
Figure 1.3 General technical system of AI technology.
Figure 1.4 Technical characteristics of new generation AI technology.
Figure 1.5 National AI strategies.
Figure 1.6 Competition pattern of AI in the future.
Figure 1.7 Relationship between data, algorithm, calculation, and artificial...
Chapter 2
Figure 2.1 Schematic diagram of the relationship between the three elements ...
Figure 2.2 Artificial intelligence ecology in satellite field.
Figure 2.3 Opportunities for the development of artificial intelligence in s...
Chapter 4
Figure 4.1 Schematic diagram of machine learning.
Figure 4.2 Deep Reinforcement Learning Framework.
Chapter 6
Figure 6.1 Structural composition of intelligent remote‐sensing satellite lo...
Chapter 7
Figure 7.1 Conceptual model of Mitola.
Figure 7.2 Structure diagram of digital channelizer.
Figure 7.3 Schematic diagram of WGS digital channelizer.
Figure 7.4 Implementation block diagram of digital channelizer based on freq...
Chapter 8
Figure 8.1 Technical composition analysis framework.
Figure 8.2 Structure of evolutionary hardware.
Figure 8.3 Classification of evolutionary hardware research.
Figure 8.4 Evolutionary hardware model.
Figure 8.5 Simplified FPGA structure.
Chapter 10
Figure 10.1 Autonomous control structure of Shenkong 1.
Guide
Cover Page
Series Page
Title Page
Copyright Page
About the Authors
Preface
Table of Contents
Begin Reading
Index
WILEY END USER LICENSE AGREEMENT
Pages
ii
iii
iv
xiii
xiv
xv
xvi
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
95
96
97
98
99
100
101
102
103
104
105
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
IEEE Press445 Hoes Lane Piscataway, NJ 08854
IEEE Press Editorial Board Sarah Spurgeon, Editor in Chief
Jón Atli Benediktsson
Behzad Razavi
Jeffrey Reed
Anjan Bose
Jim Lyke
Diomidis Spinellis
James Duncan
Hai Li
Adam Drobot
Amin Moeness
Brian Johnson
Tom Robertazzi
Desineni Subbaram Naidu
Ahmet Murat Tekalp
Intelligent Satellite Design and Implementation
Jianjun Zhang
China Academy of Space TechnologyBeijing, China
Jing Li
Beijing Institute of TechnologyBeijing, China
Copyright © 2024 by The Institute of Electrical and Electronics Engineers, Inc.All rights reserved.
Published by John Wiley & Sons, Inc., Hoboken, New Jersey.Published simultaneously in Canada.
No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, scanning, or otherwise, except as permitted under Section 107 or 108 of the 1976 United States Copyright Act, without either the prior written permission of the Publisher, or authorization through payment of the appropriate per‐copy fee to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, (978) 750‐8400, fax (978) 750‐4470, or on the web at www.copyright.com. Requests to the Publisher for permission should be addressed to the Permissions Department, John Wiley & Sons, Inc., 111 River Street, Hoboken, NJ 07030, (201) 748‐6011, fax (201) 748‐6008, or online at http://www.wiley.com/go/permission.
Trademarks: Wiley and the Wiley logo are trademarks or registered trademarks of John Wiley & Sons, Inc. and/or its affiliates in the United States and other countries and may not be used without written permission. All other trademarks are the property of their respective owners. John Wiley & Sons, Inc. is not associated with any product or vendor mentioned in this book.
Limit of Liability/Disclaimer of Warranty: While the publisher and author have used their best efforts in preparing this book, they make no representations or warranties with respect to the accuracy or completeness of the contents of this book and specifically disclaim any implied warranties of merchantability or fitness for a particular purpose. No warranty may be created or extended by sales representatives or written sales materials. The advice and strategies contained herein may not be suitable for your situation. You should consult with a professional where appropriate. Further, readers should be aware that websites listed in this work may have changed or disappeared between when this work was written and when it is read. Neither the publisher nor authors shall be liable for any loss of profit or any other commercial damages, including but not limited to special, incidental, consequential, or other damages.
For general information on our other products and services or for technical support, please contact our Customer Care Department within the United States at (800) 762‐2974, outside the United States at (317) 572‐3993 or fax (317) 572‐4002.
Wiley also publishes its books in a variety of electronic formats. Some content that appears in print may not be available in electronic formats. For more information about Wiley products, visit our web site at www.wiley.com.
Library of Congress Cataloging‐in‐Publication DataNames: Zhang, Jianjun (Writer on artificial satellites), author. | Li, Jing, 1982‐ author.Title: Intelligent satellite design and implementation / Jianjun Zhang, Jing Li.Description: Hoboken, New Jersey: Wiley, [2024] | Includes index.Identifiers: LCCN 2023028789 (print) | LCCN 2023028790 (ebook) | ISBN 9781394198955 (hardback) | ISBN 9781394198962 (adobe pdf) | ISBN 9781394198979 (epub)Subjects: LCSH: Artificial satellites. | Artificial intelligence.Classification: LCC TL796 .Z429 2024 (print) | LCC TL796 (ebook) | DDC 629.460285/63–dc23/eng/20230909LC record available at https://lccn.loc.gov/2023028789LC ebook record available at https://lccn.loc.gov/2023028790
Cover Design: WileyCover Image: © Adastra/Getty Images
About the Authors
Jianjun Zhang, PhD, ProfessorHe received PhD degree from the Institute of Optoelectronics, Chinese Academy of Sciences, in 2010. He is a professor at the Beijing Institute of Spacecraft System Engineering, China Academy of Space Technology. He is also member of the Youth Science Club of China Electronics Society, member of the Edge Computing Expert of China Electronics Society, Chairman of the “Space (Aerospace) Information Technology” Professional Committee of China Electronics Society, and member of the Satellite Application Expert Group of China Aerospace Society. He is chiefly engaged in satellite navigation system design and advanced spatial information system technology based on cognitive mechanism. He has presided over several major projects such as the National Natural Science Foundation's major research project, the final assembly fund, the 863 project, and the development project of the Science and Technology Commission of the China Academy of Space Technology. He has published more than 50 SCI/EI search papers in international journals and conferences, authorized more than 20 invention patents at home and abroad, and published 3 monographs. He won the third prize of National Defense Science and Technology Progress Award.
Jing Li, PhD, Associate Professor, SupervisorShe received PhD degree from Beijing Institute of Technology in 2011. She is an associate professor of School of Automation, Beijing Institute of Technology. She is an expert member of the “Space Information Technology” Youth Committee of China Electronics Society; her main research direction is robot environmental awareness, image detection and target tracking, and multi‐sensor information fusion. She has presided over more than 10 projects including the National Natural Science Youth Fund, the Postdoctoral Special Fund, the Key Laboratory of the Ministry of Education, and the Science and Technology Cooperation. She has published 25 academic papers (including 10 SCI papers and 15 EI papers) and the book “Image Detection and Target Tracking Technology” and has been granted 7 national invention patents as the first author and the National Science and Technology Progress Award 2 (ranked 8th). The postgraduates whom she guided have been awarded the second prize of the 14th China Graduate Electronic Design Competition, the second prize of the first China‐Russia (Industrial) Innovation Competition, and the second prize of the 14th National College Student Smart Car Competition.
Preface
Artificial Intelligence (AI) is an interdisciplinary subject developed by integrating computer science, cybernetics, information theory, neurophysiology, psychology, linguistics, philosophy, and other disciplines. It is one of the three cutting‐edge technologies in the 21st century (genetic engineering, nanoscience, AI). Through AI technology, machines can be competent for some complex tasks that usually require human intelligence to complete, thus greatly simplifying manual operations, improving production efficiency, and improving production relations. It is a very disruptive, cutting‐edge technology.
At present, the international space power led by NASA has taken space as an important stage for AI to play its role. Many space tasks that have been carried out or will be carried out have more or less adopted AI technology to improve the efficiency of related tasks. Although the current application of AI technology in space is still limited and the achievements are not outstanding enough, the power of AI has been demonstrated, and the future development direction it represents has also begun to emerge. The successful application of AI in various fields has laid a good foundation for the design of satellite systems and the development of satellite intelligence in the future.
In a series of processes such as satellite development, testing, flight control, delivery and use, the problems of the unattended space environment, the high cost of testing and maintenance, and many factors of fault problems have been puzzling scientific researchers. AI‐supporting satellite system technology is a powerful means to solve these problems and is one of the development directions of satellite system design in the future. In the future, it will not only be able to process complete information but also process incomplete information, and even intelligently supplement incomplete information, and make the processing of information and data more mature, efficient, and accurate according to the feedback system. At the same time, experience is constantly accumulated in daily operation, so that the AI system can adapt to the changing environment, gradually realize the automatic evolution mechanism, and make the AI system itself constantly learn, changing the single passive processing information into active, intelligent processing information, and even have a certain predictive ability.
With the increasing development of AI algorithms and application technologies, the next development of intelligent satellites will focus on all aspects: developing the design of onboard intelligent chips to lay the hardware foundation for satellite intelligence. Develop satellite system design based on AI to realize a processing platform that can meet the flexible expansion of multiple tasks and support the flexible reconfiguration of system resources in case of failure. Develop the on‐orbit fault detection and maintenance technology based on AI to realize the monitoring of satellite on‐orbit status. Carry out research on satellite intelligent control technology based on AI, and realize the application of real‐time intelligent autonomous attitude control, intelligent autonomous GNC, and intelligent information technology in aerospace control systems, platforms, and payloads. Carry out research on satellite‐ground integration technology based on AI and build a satellite‐ground integration satellite platform. Finally, combined with intelligent learning algorithm, the intelligent task of satellite platform is realized.
Jianjun Zhang
China Academy of Space TechnologyBeijing, China
Jing Li
Beijing Institute of TechnologyBeijing, China
1Development of Artificial Intelligence
1.1 The Concept and Evolution of Artificial Intelligence
1.1.1 The Concept of Artificial Intelligence
Artificial intelligence (AI), also known as machine intelligence, refers to the intelligence represented by machines made by people. Generally, AI refers to human intelligence technology realized by means of various ordinary computer programs. The definition in the general textbook is “the research and design of intelligent agent,” which refers to a system that can observe the surrounding environment and make actions to achieve goals [1, 2].
The definition of AI can be divided into two parts, namely, “artificial” and “intelligence.” “Artificial” is easier to understand and less controversial. Sometimes we consider what humans can do and create, or whether a person's own level of intelligence is high enough to create AI, and so on. But to sum up an “artificial system” is an artificial system in the general sense. There are many questions about what “intelligence” is. This involves other issues such as consciousness, self and mind, including the unconscious mind. The only intelligence that people know is their own intelligence, which is a widely accepted view. But our understanding of our own intelligence is very limited, and our understanding of the necessary elements of human intelligence is also very limited, so it is difficult to define what “intelligence” is made by “artificial.” Therefore, the research on AI often involves the research on human intelligence itself. Other intelligence with animals or other artificial systems is also generally considered as a research topic related to AI.
A popular definition of AI, as well as an earlier definition in this field, was put forward by John McCarthy of the Massachusetts Institute of Technology at the Dartmouth Conference in 1956: AI is to make the behavior of machines look like that of human beings. But this definition seems to ignore the possibility of strong AI. Another definition is that AI is the intelligence represented by artificial machines. In general, the current definition of AI can be divided into four categories, namely, machines “think like people,” “move like people,” “think rationally,” and “act rationally.” Here, “action” should be broadly understood as the decision to take action or specify action, rather than physical action.
Strong AI believes that it is possible to produce intelligent machines that can really reason and solve problems, and such machines will be considered as perceptual and self‐conscious. There are two types of strong AI:
Human‐like AI, that is, the thinking and reasoning of machines, is like human thinking.
Nonhuman AI, that is, machines produce perception and consciousness completely different from human beings and use reasoning methods completely different from human beings.
The term “strong artificial intelligence” was originally created by John Rogers Hiller for computers and other information‐processing machines. Its definition is: strong AI holds that computers are not only a tool for studying human thinking. On the contrary, as long as it runs properly, the computer itself is thinking. The debate on strong AI is different from the debate on monism and dualism in a broader sense. The main point of the argument is: if the only working principle of a machine is to convert encoded data, then is the machine thinking? Hiller thought it was impossible. He gave an example of a Chinese room to illustrate that if the machine only converts data, and the data itself is a coding representation of some things, then without understanding the correspondence between this coding and the actual things, the machine cannot have any understanding of the data it processes. Based on this argument, Hiller believes that even if a machine passes the Turing test, it does not necessarily mean it is really thinking and conscious like a person. There are also philosophers who hold different views. Daniel Dennett believes in his book Consciousness Explained that man is just a machine with a soul. Why do we think: “Man can have intelligence, but ordinary machines can't?” He believes that it is possible to have thinking and consciousness when data is transferred to machines like the above.
The weak AI point of view believes that it is impossible to produce intelligent machines that can really reason and solve problems. These machines just look intelligent, but they do not really have intelligence, nor do they have autonomous consciousness. Weak AI came into being when compared with strong AI because the research on AI was at a standstill for a time, and it began to change and go far ahead until the artificial neural network (ANN) had a strong computing ability to simulate. In terms of the current research field of AI, researchers have created a large number of machines that look like intelligence, and obtained quite fruitful theoretical and substantive results, such as the Eureqa computer program developed by Cornell University Professor Hod Lipson and his doctoral student Michael Schmidt in 2009, as long as some data are given, This computer program can deduce the Newtonian mechanics formula that Newton spent years of research to discover in only a few hours, which is equivalent to rediscovering the Newtonian mechanics formula in only a few hours. This computer program can also be used to study many scientific problems in other fields. These so‐called weak AI have made great progress under the development of neural networks, but there is no clear conclusion on how to integrate them into strong AI.
Some philosophers believe that if weak AI is realizable, then strong AI is also realizable. For example, Simon Blackburn said in his introductory philosophy textbook Think that a person's seemingly “intelligent” action does not really mean that the person is really intelligent. I can never know whether another person is really intelligent like me or whether they just look intelligent. Based on this argument, since weak AI believes that it can make the machine look intelligent, it cannot completely deny that the machine is really intelligent. Blackburn believes that this is a subjective issue. It should be pointed out that weak AI is not completely opposite to strong AI, that is, even if strong AI is possible, weak AI is still meaningful. At least, the things that computers can do today, such as arithmetic operations, were considered to be in great need of intelligence more than 100 years ago. And even if strong AI is proven to be possible, it does not mean that strong AI will be developed.
In a word, AI is an interdisciplinary subject, belonging to the intersection of natural science and social science. The disciplines involved include physics, philosophy and cognitive science, logic, mathematics and statistics, psychology, computer science, cybernetics, determinism, uncertainty principle, sociology, criminology, and intelligent crime. The research on AI is highly technical and professional, and each branch field is deep and different, so it covers a wide range. The core issues of AI include the ability to construct reasoning, knowledge, planning, learning, communication, perception, movement, and operation of objects that are similar to or even beyond human beings. Strong AI is still the long‐term goal in this field.
There are different understandings about the definition of AI from different perspectives, and there is no unified definition at present. Here are some definitions with high acceptance:
In 1956, scientists such as John McCarthy of Stanford University, Marvin Minsky of Massachusetts Institute of Technology, Herbert Simon and Allen Newell of Carnegie Mellon University, Claude Shannon of Bell Laboratory, and other scientists first established the concept of “artificial intelligence” in Dartmouth College of the United States, that is to say, let machines recognize, think, and learn like people, that is, use computers to simulate human learning and other aspects of intelligence. At the same time, seven typical task directions have been established: machine theorem proving, machine translation, expert system, game, pattern recognition, learning, robot, and intelligent control.
In 1981, A. Barr and E.A. Feigenbaum proposed from the perspective of computer science: “Artificial intelligence is a branch of computer science. It is concerned with the design of intelligent computer systems, which have intelligent characteristics associated with human behavior, such as understanding language, learning, reasoning, problem solving, etc.”
In 1983, Elaine Rich proposed that “Artificial intelligence is to study how to use computers to simulate human brain to engage in reasoning, planning, design, thinking, learning and other thinking activities, and to solve complex problems that are still considered to need to be handled by experts.”
In 1987, Michael R. Genesereth and Nils J. Nilsson pointed out that “AI is the science of studying intelligent behavior, and its ultimate purpose is to establish a theory on the behavior of natural intelligence entities and guide the creation of artificial products with intelligent behavior.”
Wikipedia defines “AI is the intelligence displayed by machines,” that is, as long as a certain machine has some or some “intelligence” characteristics or performance, it should be counted as “AI.” The Encyclopedia Britannica defines “AI is the ability of a digital computer or a robot controlled by a digital computer to perform some tasks that intelligent organisms only have.” Baidu Encyclopedia defines AI as “a new technological science that studies and develops theories, methods, technologies, and application systems for simulating, extending and extending human intelligence,” regards it as a branch of computer science, and points out that its research includes robots, language recognition, image recognition, natural language processing, and expert systems.
According to the White Paper on the Standardization of Artificial Intelligence (2018), “Artificial intelligence is a theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. Artificial intelligence is the engineering of knowledge, which is the process that machines imitate human beings to use knowledge to complete certain behaviors.”
The US Defense Authorization Act of FY 2019 defines AI as all systems that can “act rationally” in a changing and unpredictable environment without sufficient human supervision or can learn from experience and use data to improve performance.
To sum up, AI allows computer/machine to simulate, extend, and expand human intelligence, so that the system can cope with changing and unpredictable environments without human supervision, deal with complex problems through learning and experience, and achieve performance improvement [2, 3].
1.1.2 Evolution of Artificial Intelligence
AI was born in the 1950s. Over the past 60 years, it has experienced three waves (Figure 1.1) and has formed such theoretical schools as semiotics, connectionism, and behaviorism. The first wave (1956–1976) was dominated by semiotics, and its main achievements were machine reasoning, expert system, and knowledge engineering; the second wave (1976–2006), dominated by the connectionist school, produced neural networks, machine learning, and other achievements; the third wave (from 2006 to now), led by the connecting school, achieved great success in deep neural network and deep reinforcement learning, and AI entered the stage of commercial development. The behavioral school theory has not played a leading role in the third wave so far. Its core is adaptive control, learning control, evolutionary computing, and distributed intelligence, which are the basis of modern control theory [3].
In the three waves of AI development, the key points in technology and application breakthrough are shown in Figure 1.2.
Notably, the behavioral school theory has not played a leading role in the third wave so far. Its core is adaptive control, learning control, evolutionary computing, and distributed intelligence, which is the basis of modern control theory. In recent years, different voices have emerged in the development of AI. They believe that AI technology based on the connectionist school theory has a huge gap in solving problems that cannot be reasoned and counted and cannot achieve strong AI with creative thinking. They believe that the symbolic school and behavioral school theory pay more attention to the abstraction of human higher intelligence and more attention to adaptation and evolution; the realization of strong AI in the future may need to find another way.
Figure 1.1 Three waves of AI.
Figure 1.2 Key points of AI technology development.
1.2 The Current Scope and Technical Framework of Artificial Intelligence
1.2.1 Technical Scope
Compared with the early AI, the new generation AI is under the guidance of the new information environment, massive database, and continuously evolving and enriching strategic objectives, relying on the two basic platforms of cloud computing and big data and the three general technologies of machine learning, pattern recognition, and human–computer interaction and taking the new computing architecture, general AI, and open source ecosystem as the main guidance, and continues to build and improve the technical framework system. As a result of multi‐disciplinary intersections and universal technology, AI technology has formed a complex network of technology systems together with related downstream technologies and applications. At present, the network is in its infancy, but it is still in a dynamic state of rapid renewal and drastic changes [4].
1.2.2 Technical Framework
According to the description in the White Paper on the Development of New Generation AI (2017), the current AI technical framework is mainly composed of three parts: basic layer, technical layer, and application layer, as shown in Figure 1.3:
Basic layer
The basic layer mainly includes big data, smart sensors, smart chips, and algorithm models. Among them, smart sensors and smart chips belong to basic hardware, and algorithm models belong to core software [5].
Technical level
The technical level mainly includes pattern recognition, autonomous planning, intelligent decision‐making, autonomous control, human–machine cooperation, group intelligence, etc.
Application layer
The application layer mainly includes intelligent agents (robots, unmanned driving, intelligent search, and unmanned aerial vehicles) and industrial applications (finance, medical, security, education, human settlements, etc.).
Figure 1.3 General technical system of AI technology.
1.2.3 Technical Features
At present, the development of AI is based on the theory of connectivity school. The main driving factors are explosive growth of data, continuous improvement of computing power, continuous optimization of algorithm models, and deep coupling of capital and technology. The main technologies include computer vision, machine learning, natural language processing, robot technology, speech recognition, etc. At present, intelligent medical, finance, education, transportation, security, home manufacturing, and other fields have been widely used and developed rapidly in unmanned driving, intelligent robots, and others. Among them, deep learning is a typical representative of the development of the new generation of AI, but “deep learning” is not a synonym for AI [6, 7].
Driven by data, computing power, algorithm model, and multiple applications, AI is evolving from auxiliary equipment and tools to assist and partner for collaborative interaction and is more closely integrated into human production and life (Figure 1.4):
Big data has become the cornerstone of the sustained and rapid development of AI.
Text, image, voice, and other information realize cross‐media interaction.
Network‐based swarm intelligence technology has begun to sprout.
Autonomous intelligent system has become an emerging development direction.
Human–machine collaboration is giving birth to a new type of hybrid intelligence.
Figure 1.4 Technical characteristics of new generation AI technology.
1.3 The Overall Development Trend of Artificial Intelligence
1.3.1 Current Development Trend
Artificial intelligence technology, as an important factor leading industry disruption and technological change, has achieved explosive growth in all fields around the world.
In the past decade, intelligent algorithms represented by deep learning and intensive learning have promoted AI to continue to exert its power in cutting‐edge technologies and applications, and giant companies such as Google and IBM have made major breakthroughs in search engines, human–computer games, intelligent medical care, unmanned driving, and other fields. By December 2018, more than 5000 AI startups had been born and flourished in reading and writing assistant, financial industry, enterprise management, advanced manufacturing, and other industries, which had a profound impact on various fields [8–10].
The major developed countries in the world regard the development of AI as a major strategy to enhance national competitiveness and maintain national security.
The United States was the first country to release a national strategy for AI. In 2016, it released the National Strategic Plan for AI Research and Development. In the next two years, a total of 15 countries and regions released their AI strategies and accelerated their planning and layout in the field of AI (Figure 1.5). Among them, thanks to decades of federal research funds, industrial production, academic research, and the inflow of foreign talents, the United States has been leading the global wave of AI development in basic theory, software and hardware, talents, enterprises, and other aspects.
In many aspects, AI is far from mature application, and there are still technical bottlenecks.
At present, it is generally believed that AI has surpassed human beings in large‐scale image recognition. Although some progress has been made in machine translation, it is not close to the ideal level, while there is a big gap in chat conversation. In terms of driverless driving, the current commercial autonomous vehicle is auxiliary driving, and the real autonomous vehicle is still in the development and testing stage, in which dealing with emergency and abnormal traffic conditions is the difficulty encountered by AI. Similarly, in speech recognition, although AI recognition is close to humans in the experimental environment, in reality, especially in the presence of environmental interference, the recognition rate of AI is actually unsatisfactory, and it often makes some common sense mistakes that humans cannot make. Some experts even believe that AI technology will not make a major breakthrough in 20 years, because there are not many new topics in the field of AI research done by the current academic community. Even if scientists work hard, it will take about 20 years to accumulate a theoretical basis that makes people feel very excited and surprised.
AI will accelerate the intelligent transformation and upgrading of traditional industries.
In the future, with the continuous improvement of AI capabilities, the integration of AI and traditional industries will continue to deepen, driving the transformation and upgrading of traditional industries to intelligence and at the same time creating huge benefits for society. According to the research report, by 2025, the market scale of AI in agricultural applications will reach 20 billion dollars, and the financial industry will save and generate 34–43 billion dollars in costs and new market opportunities, and the annual expenditure in the medical field will be reduced by 54 billion dollars.
Human–computer cooperation will become the main direction of AI commercialization.
With the complementary nature between humans and AI systems, the collaborative interaction between humans and AI systems will become the main direction of AI commercialization. Although fully autonomous AI systems will play an important role in underwater or deep space exploration and other applications, AI systems cannot completely replace humans in the short term in disaster recovery, medical diagnosis, and other applications.
Figure 1.5 National AI strategies.
Human–machine cooperation can be divided into three ways, namely, joint execution, auxiliary execution, and alternative execution:
Joint execution
AI system positioning: perform peripheral tasks supporting human decision makers.
Typical applications: short‐term or long‐term memory retrieval and prediction tasks.
Auxiliary execution
AI system positioning: when human beings need help, the artificial intelligence system performs complex monitoring functions.
Typical applications: ground proximity alarm system, decision‐making, and automatic medical diagnosis in aircraft.
Alternative execution
AI system positioning: AI systems perform tasks with very limited capabilities for humans.
Typical applications: complex mathematical operations, dynamic system control and guidance in controversial operating environment, automatic system control in hazardous or toxic environment, nuclear reactor control room, and other rapid response systems.
AI will present a competitive pattern of leading platform plus scenario application.
Under the trend of AI platformization, AI will present a competitive pattern of several leading platforms and extensive scene applications in the future, and ecological builders will become the most important model among them. As shown in Figure 1.6, the future AI competition pattern will mainly present five modes.
Mode 1: Ecological construction – take the whole industrial chain ecology and scenario application as a breakthrough. Take Internet companies as the main body, mainly invest in infrastructure and technology for a long time. At the same time, scenario applications will be used as traffic entry, accumulate applications, and become the leading application platform. It will become the builder of AI ecosystem.
Mode 2: Technical algorithm‐driven – take technical layer and scene application as a breakthrough. Take software companies as the main body, deeply cultivate algorithm platform and general technology platform, and gradually establish application platform with scene application as the traffic entrance.
Mode 3: Application focus – scene application. Based on the scenario or industry data, many segmented scenario applications are developed mainly for startups and traditional industry companies.
Figure 1.6 Competition pattern of AI in the future.
Mode 4: Vertical field first – killer application and gradually builds vertical field ecology. It is mainly a pioneer in the vertical field. It relies on killer applications to accumulate a large number of users and data in the vertical field, and deeply cultivates the common technologies and algorithms in the field, becoming a subverter in the vertical field.
Mode 5: Infrastructure provision – starting from infrastructure and expanding to the downstream of the industrial chain. Take chip or hardware and other infrastructure companies as the main part, start from infrastructure, improve technical capabilities, and expand to the upstream of data, algorithms, and other industrial chains.
1.3.2 Technical Development Trend
At present, AI, represented by deep learning and big data, has made amazing progress in the fields of image classification, speech recognition, visual understanding, and machine translation. However, deep learning relies on annotated data and lacks the ability to express logical reasoning and causal relationships. It is difficult to deal with tasks with complex spatiotemporal correlations and cannot achieve the goal of strong AI. In the future, AI technology needs to continue to break through [10–12].
From deep learning to brain intelligence algorithm
In the aspect of algorithms, the basis of AI technology is developed according to two main lines: deep learning improvement and new algorithms. Advanced in‐depth learning focuses on breaking through the methods of adaptive learning, autonomous learning, zero‐data learning, unsupervised learning, migration learning, etc., to achieve AI with high reliability, high interpretability, and strong generalization ability. Second, the academic community continues to explore new algorithms, and brain‐like intelligent algorithms have become a frontier hotspot, focusing on breaking through brain‐like information coding, processing, memory, learning, association, reasoning, and other theories, forming brain‐like complex systems and brain‐like control and other theories and methods, and establishing new models of large‐scale brain‐like intelligent computing and brain‐inspired cognitive computing models [12, 13].
From dedicated intelligence to general intelligence
With the continuous development of science and technology and the deep transformation of social structure, the problems facing human beings are highly complex. There is an urgent need for a wide range, high integration, and strong adaptability of general wisdom in professional fields such as game, identification, control, and prediction, which significantly improves the ability of human beings to read, manage, and reorganize knowledge. General AI has the characteristics of reducing dependence on domain knowledge, improving the applicability of processing tasks, and achieving the correction of machine‐independent cognition. It has the ability to process multiple types of tasks and adapt to unexpected situations. Its substantive progress will truly start the prelude of intelligent revolution, highly integrate with the existing physical and information world, and profoundly affect all aspects of social and economic development.
From machine intelligence to human–machine hybrid intelligence
Machine intelligence (AI) and human intelligence have their respective strengths and need to learn from each other and complement each other. Integrating multiple intelligence models, human–machine coexistence will become the new normal of the future society. It will focus on breaking through the theory of human–machine win–win situation understanding and decision learning, intuitive reasoning and causal model, memory and knowledge evolution, and realize hybrid enhanced intelligence that learning and thinking are close to or exceed the level of human intelligence.
From single agent intelligence to group intelligence
Group intelligence originates from the research on the group behavior (including information transmission, collective decision‐making, etc.) of social insects represented by ant colonies, bee colonies, etc. The group has self‐organization, division of labor, and coordination. Group intelligence refers to the collective intelligent behavior of a system composed of individual units through interaction between each other or with the environment, with the characteristics of intelligent emergence. Science believes that the era of group intelligence based on networks is coming. Group intelligence will focus on breaking through the theories and methods of organization, emergence and learning of group intelligence, establishing expressive and computable group intelligence incentive algorithms and models, and forming group intelligence capabilities based on a network environment [14, 15].
Weak AI will gradually transform into strong AI
AI can be divided into three stages: weak AI, strong AI, and super AI. The level of AI in the three stages has been continuously improved. Weak AI refers to human ability in some fields; strong AI refers to having human capabilities in all fields, being able to compete with human beings in all aspects, and being unable to simply distinguish between human beings and machines; super AI means that it can surpass human beings in all fields, and can surpass human beings in the fields of innovation and creative creation to solve any problems that cannot be solved by human beings.
From the current development level of AI, AI is still a weak AI based on specific application fields, such as image recognition, speech recognition, and other biometric analysis, such as intelligent search, intelligent recommendation, intelligent sorting, and other intelligent algorithms. When it comes to vertical industries, AI mostly assists human beings in their work in the role of assistant, such as the current intelligent investment advisor and autonomous vehicle, but the AI that completely gets rid of human beings and can reach or even surpass human beings in the true sense cannot be realized. In the future, with the substantial growth of computing power and data volume and the improvement of algorithms, weak AI will gradually transform into strong AI, and machine intelligence will advance from perception, memory, and storage to cognition, autonomous learning, decision‐making, and execution.
Data, algorithm, and computing power are the three carriages of AI
In recent years, with the increase of the data volume index, the algorithm theory continues to be updated iteratively, and the computing ability continues to be enhanced, which jointly promotes the accelerated development of AI. The relationship between the three and AI is shown in Figure 1.7.
With the rapid development of the Internet, not only the standard training set and test set data are increasing but also the massive and high‐quality application scenario data are becoming increasingly rich. Real‐time, massive, multi‐sources, and multi‐types of data can describe reality more closely from different angles, and machine learning algorithms can be used to mine multi‐level associations between data, laying the foundation of data sources for AI applications.
Figure 1.7 Relationship between data, algorithm, calculation, and artificial intelligence.
The development of algorithms, especially the paper published by Professor Geoffrey Hinton in 2006, has started the wave of deep learning in academia and industry. The deep learning algorithm represented by ANN has become the core engine of AI applications.
AI has a high demand for computing power. With the rapid development of cloud computing technology and chip processing power, thousands of machines can be used for parallel computing. In particular, the development of tensor processing unit (TPU), graphics processing unit (GPU), field programmable gate array (FPGA), and special chips for AI has laid the foundation for implementing AI computing power, making the application of AI similar to human's deep neural network algorithm model a reality.
