The 10th International Symposium on the Integration of Computational Communication and Control Science was successfully held, and experts such as Academician Li Guojie and Academician Sun Ninghui disc

With the rapid development of the new generation of information technology, intelligent systems represented by intelligent robots, unmanned systems, industrial Internet, artificial intelligence models, etc. are gradually leading a new round of major changes in social production and life, and becoming an important position for national strategic science and technology competition under the unprecedented change in the world in a century. Computing (IT), communication (CT), and control (OT), as the three cornerstones of intelligent systems, are gradually showing a trend of deep cross integration.

In order to accelerate the solution of the key bottleneck problems of intelligent systems such as robots, unmanned systems and industrial Internet in China, and improve the innovation capability of key core technologies, the 10th International Academic Seminar on the Integration of Computing, Communication and Control Sciences and the Guangdong Hong Kong Macao Greater Bay Area Intelligent System Summit Forum in 2023 were held in Zengcheng, Guangzhou. Academicians Li Guojie and Sun Ninghui of the Chinese Academy of Engineering led a team from computing institutes such as Sinsegye to participate in this summit.

 Li Xiaoning, Senior Product Director and General Manager of Shanghai Company at Sinsegye, Liu Shengfu, Senior Vice President of Sinsegye, Academician Li Guojie of the Chinese Academy of Engineering

The International Symposium on the Integration of Communication Networks and Computing Science was led and organized by Academician Li Guojie of the Chinese Academy of Engineering in 2014. It has been successfully held for nine sessions and has had a profound impact both domestically and internationally. However, OICT fusion for intelligent systems still faces numerous challenges. What theory can guide OICT fusion for intelligent systems? Where is the breakthrough point for OICT fusion in intelligent systems? Academician Li Guojie pointed out in his opening speech that regarding the deep integration of computing and control, computing technology needs to further sink, including to the edge and terminals. In a sense, we need to take a path of computerization. The birth time of the two disciplines of computation and control is almost the same, both in the 1940s. In the early days, the combination of these two fields was very obvious, for example, in China, control is called automation. In fact, scholars at that time mainly engaged in research on pattern recognition and artificial intelligence, which was not much different from computers. However, the difficulty of combining automation with OT technology is significant.

The birth of artificial intelligence machines

In the early integration of computing and control, a typical technology was the programmable logic controller (PLC). From a computer perspective, PLC is a small embedded computer system that integrates certain control functions. However, with the increasing demand for computing power in intelligence, PLCs are no longer able to meet the demand. Therefore, an industrial control computer with high computing power was added to the industrial site as the upper computer, while PLC and motion controller were used as the lower computers. The upper computer was responsible for human-computer interaction, scheduling decision-making, and monitoring, while the lower computer was responsible for detection and execution. However, this architecture still handles relatively simple control and cannot be considered a true fusion of computation and control. Currently, many intelligent applications include perception, judgment, decision-making, and action, which we call the OODA loop. This loop puts higher demands on the computing platform of on-site software and algorithms, similar to the evolution of feature phones into smartphones, traditional cars into intelligent cars, and industrial automation needs to develop towards computerization. In other words, we need a computing platform similar to the iPhone This multi in one product, which balances the computing power and real-time capability of the terminal, is called an industrial intelligent computer, abbreviated as an AI machine. At the same time, it requires an operating system, a real-time domain control domain, and a non real time domain computing domain, so that the architecture of computing and control implemented on different machines needs to be transformed into control and communication between processes within one machine. This will eliminate the so-called cross machine and cross operating system communication between the upper computer and lower computer. Those who have worked with computers know that as long as cross operating system communication is done, it will be very slow.

Based on the above ideas, the Intelligent Computer Research Center of the Institute of Computing Technology, Chinese Academy of Sciences has conducted research in this area. They have done some work at the operating system level, application software level, and even at the chip level. They have developed a chip called Dadu, which is used for real-time control. In 2022, with the organizational support of the computing institute, a company called Sinsegye was established to produce artificial intelligence machines, providing industrial automation products based on PC technology and software defined technology. It has raised funds five times within a year and has launched five artificial intelligence machine products. This year's product revenue may exceed expectations, reaching the forefront of the industry and being promoted to enterprises including Huawei. Sinsegye has taken a small step on the road of computer and control. The promotion of deep learning models for artificial intelligence technology has provided more diverse attempts and broader and stronger solutions for AI machines, which helps to resolve the fragmentation of OT technology and provides new opportunities for intelligent manufacturing and industrial intelligence.

Discussion on the Integration of Computing, Communication, and Control

Subsequently, Academician Sun Ninghui shared his summarized approach and methods in the report: we divide the information age into three stages: 1.0 is the era of computers, which connects these computers into a new space through computer networks; 2.0 is an era of Internet, connecting computers and people into a binary computing era; These two eras have created a brand new virtual digital space for us. So in the era of 3.0, we need to add the physical world and deeply integrate the data space we construct with the existing physical space or the artificial physical space. This is the background that our three disciplines need to deeply integrate, and it is an example of their collaborative work. In the era of 1.0, computing, communication, and control developed independently and were not yet fully integrated with each other. Only in the early days when computer science and control science were just beginning to emerge did there be some integration. So in the 2.0 era, we found that IT and CT are closely integrated, so we often look at this industry, sometimes called IT industry, sometimes called ICT industry, so these two are inseparable. In the 3.0 era, there is an industry called OICT, and these three things are inseparable from each other. Whether it is a computing system, a communication system, or a control system, these three things are coupled together and cannot be separated. From a computational perspective, there are three ways of integration, one is called system level integration, the other is the generalization or computerization of computer technology, and the other is that computation also plays a role in improving communication performance. In fact, this fusion was first proposed by Academician Gao Wen when he started working at the Institute of Computing. He introduced the concept of "computing computer", which was called "communication". At that time, the network was not as good as it is today, and our terminals were very heavy. When he mentioned the concept of "computing computer", he turned this "fat" terminal into a "thin" terminal, but now we are accustomed to calling it a network computer. He actually converts the power of the network into a decrease in computing power, so he doesn't need too much computing power. This is the earliest prototype of fusion. Everyone knows about smartphones, and we consider them to be a fusion of computing and communication. Below is a feature phone that doesn't do much computing, while above it, both hardware and software are specialized in computer science, CPU、GPU、 Operating systems, graphics rendering, and general-purpose programming environments are all concepts of computers, and only at the bottom are concepts of communication. OODA is a way of intelligent systems. In the military field, it is to decompose a relatively complex intelligent process into four steps: perception, judgment, decision-making, and execution. Then, multiple iterative loops can be used to decompose a complex system into a superposition of simple systems. So the idea of OODA loop is to map an intelligent process into a computable one, where each step can be computed and connected through algorithms. This is very similar to the closed-loop iteration principle of control theory. So many intelligent systems in real life, whether military, industrial, or agricultural, are like this. Except for sensors, its actuators all interact with the physical world. To construct such a system, its programming interface is not yet universal. Everyone can create such a system, but there is no unified change interface.

The Approach of Artificial Intelligence Machines

Academician Sun Ninghui said about artificial intelligence machines: When it comes to making intelligent machines, one of our application scenarios is a device in the field of national defense. The scenario of this device is very fragmented and requires a universal platform to support it. This is very similar to the demand in the industrial sector, which is also highly fragmented. Therefore, we unfolded the O-ring of this industry and found that we are using industrial infrastructure. If we add the need for intelligence, it will become very fragmented and difficult to support complex industrial scenarios. Each layer has its shortcomings. If it is only for running simple control functions, then traditional PLC has better controllability and has a mature development time. But if we add intelligence to this, we will find that we have no way to start. So we need to use a computer like approach to provide sufficient support from the bottom layer, without using dedicated MCUs, chips, or real-time operating systems. Instead, we can use Haiguang's CPU, Cambrian's acceleration card, and then use general-purpose platforms such as Linux on top to modify it to support virtualization. Due to virtualization, real-time performance cannot be guaranteed. Industry requires good real-time performance, and all operations require a fixed time and resources need to be immediately guaranteed. So we need an architecture that can ensure real-time system operation, and the original computer systems did not consider this issue. When dealing with this situation, we need to use the idea of integrating computation and control. This is equivalent to intelligent applications that can be implemented on one platform. This is also an interesting but immature idea. We took this opportunity to discuss with experts in the control field and found that industrial Internet applications are also fragmented, with many and various applications, but we do not have a unified digital way. Therefore, we need to add an abstraction layer on top and name it the computable manufacturing industrial operating system. Our idea is to add an abstraction layer between the complex and diverse hardware at the bottom and the diverse software at the top, based on the concept of computer operating systems.  

With the theme of "cross integration of computing, communication and control to promote the development of intelligent system technology", the seminar focused on the perception communication computing integration of intelligent autonomous unmanned systems, the industrial digital architecture of computing, communication and control integration, the method of computing, communication and control integration for industrial Internet, and the artificial intelligence paradigm of computing communication control integration, as well as related extension issues. It carried out in-depth discussions in various forms, such as academician reports, keynote reports of the conference, and invited reports of sub venues, to jointly provide suggestions for scientific and technological innovation and achievement transformation in the field of Industry 4.0 and computing manufacturing. As one of the representative enterprises of Industry 4.0 in China, Sinsegye will further realize industrial technologies such as domestically produced chips, domestically produced virtualization dual domain operating systems, and domestically produced industrial software, build an independent and controllable industrial infrastructure with "industrial intelligence machines" as the core, overtake the next generation of computers and control machines in a new form, solve the bottleneck problem of industry, help achieve the transformation and upgrading of Chinese manufacturing to Chinese intelligent manufacturing, challenge international competitors, promote national industrial development, and lead the new era of computable manufacturing.