In the "5g and network development strategy seminar" held by the future mobile communication forum, Yu Xiaohu, a professor of Southeast University, deputy chief engineer of broadband mobile communication National Science and technology major special project, and Secretary General of the future mobile communication forum, pointed out in his speech that for researchers in the field of communication, improving the performance of communication system based on AI technology is one of the research hotspots in recent years.
You Xiaohu said that within the industry, this way is called the endogenous intelligence of communication network. Its scientific problems can be attributed to the research of the operation mechanism of data-driven communication network. In fact, it is a kind of technology that analyzes and infers a large number of available data in the network, and then adjusts the nodes, the state of transmission link or resources in time. Strictly speaking, it is a kind of data-driven analysis, inference and decision-making technology.
According to the introduction, the most important basic theoretical support behind it is Bayesian theory, that is, based on the observed data, as well as the prior model, to infer and make decisions on the results of the data, which also includes maximum likelihood, maximum posterior and various suboptimal algorithms, etc., which can be used for the evaluation of the limit performance of the communication system.
According to Yu Xiaohu, data plays an important role in AI + 5g research. But "data validity" is a prerequisite, which is more important than the "learning" process. Data validity includes accuracy and identifiability, ergodicity and data missing, which needs to be tested before "learning".
He further explained that the data in the communication system usually has congenital defects in ergodicity (such as Gaussian distribution), and without human intervention, it often leads to low efficiency of "learning". At the same time, the data in communication system is often scarce (such as the lack of data corresponding to system crash in network optimization), which needs to be extrapolated and supplemented by simulation.
"The use of prior knowledge for learning depends on its own reliability. By making full use of prior knowledge, we can improve the efficiency and convergence rate of "learning", reduce the complexity of calculation and meet the real-time requirements of communication system. "
You Xiaohu pointed out that data is the basic link for AI to use in 5g. In reality, the problem we usually encounter is the lack of real data, we can only use simulation data to study, which makes our research and practice often divorced.
In order to solve this problem, Zijinshan laboratory has built PML 5g + netdata platform, which is composed of 5g core network, access network and terminal. It is a comprehensive test platform that comprehensively supports the evolution from 5g to 6G, and provides system level large-scale test verification for intelligent communication technology.
According to the introduction, pml5g + netdata platform can realize full data collection of the whole network by means of hard and soft mining and road test. For wireless test data, the platform collects air port signaling data through RCU and open port test cell phone, including PHY / MAC / RLC / PDCP / RRC layers. The soft mining method is used to collect the data of UU / xn / x2 / E1 / F1 / MR from the GNB. Core network data adopts mirror hard mining core network control surface and user surface data. Network management data, collect network management data from the north interface. Firewall data collection logs and other data.
You Xiaohu said that after the basic link of data collection was solved, the next step was data storage. After solving the problem of data storage, the laboratory will further study the problem of data modeling and the basic framework of AI. On this basis, the whole network is finally applied with big data to support services and network planning and optimization. "We want to form an open platform for the industry and academia to use together," he stressed.
-About future Forum-
Future forum is a non-profit international community organization jointly established by well-known domestic and foreign mobile communication operation enterprises, equipment manufacturing enterprises, scientific research institutions, colleges and universities, with the support of the national development and Reform Commission, the Ministry of science and technology, the Ministry of industry and information technology and the National Natural Science Foundation of China.
In order to promote the technology exchange and information exchange in the field of mobile communication in the future, strengthen the international technical cooperation and exchange, and jointly develop the mainstream technology of mobile communication in the future, this forum will strive to build and develop an open, neutral, equal, fair and international information exchange and technical cooperation platform that can serve the research and development of mobile communication technology in the future.