Photo courtesy of vision china
A very important feature of 5G is that at the protocol layer, the Internet and communication networks have truly integrated. Therefore, many of our algorithms in the computer field and technologies in the Internet field can be more easily implemented in the 5G network.
Dean of Tsinghua University Intelligent Industry Research Institute, Zhang Yaqin
"Ah, Wuhuan, you have one more ring than Sihuan." Many people have a similar understanding of 5G: 5G is 1G more than 4G, movies download in seconds, and video on demand is smooth … …
In the latest issue of Telecommunications Science, an article titled "The Next Decade of Communication Artificial Intelligence" shows people a completely different prospect of 5G communication applications: the future communication network can be highly autonomous. Through deep integration with artificial intelligence, 5G communication can realize "three selves", that is, parameters are self-configured, performance is self-optimized, and faults are self-cured.
On May 4th, Zhang Yaqin, one of the co-authors of the paper, an academician of the American Academy of Arts and Sciences and president of Tsinghua University Intelligent Industry Research Institute, said in an interview with the Science and Technology Daily reporter that the communication industry is moving towards 5G, and it will soon move towards 6G. Just watch the video faster and the signal is better. That’s not 5G.
What is the real 5G? Why do 5G networks "marry" with artificial intelligence?
Only with artificial intelligence can we break through these bottlenecks.
"The combination of 5G and artificial intelligence is an urgent matter." Wei Leping, one of the co-authors of the paper, deputy director of the Science and Technology Committee of the Ministry of Industry and Information Technology and director of the Science and Technology Committee of China Telecom Group Company, is most concerned about the energy consumption of 5G.
Compared with a 4G base station, a 5G base station consumes 3.5— of energy under half load. Four times. "If the value generated by 5G is not enough to pay the electricity bill, it will become a bottleneck for the widespread landing of 5 G." Wei Leping said that the optimization ability of artificial intelligence may solve this problem, so relevant departments, enterprises and other institutions need to dig, standardize and promote with a sense of urgency.
In addition to data limit and energy consumption bottleneck, artificial intelligence is also urgently needed for the management of huge networks.
"The 4G communication network system is enormous, and the types of networks are also very complicated. Up to now, I haven’t counted how many networks there are. Therefore, it is very difficult to manage these networks, not to mention organic integration and collaborative improvement. " Wei Leping said that the situation of 5G is more complicated, and it is even more impossible to manage such complicated resources manually. Therefore, it is very necessary for artificial intelligence to intervene, put forward early warning and give pre-judgment, and greatly improve the management efficiency of 5G networks.
In the past 10 years, they have "danced in chains"
As early as a decade ago, the communication industry has already seen the potential value of machine learning and artificial intelligence.
"In the beginning, some foreign operators were full of confidence in realizing the self-organization, self-optimization and self-healing of communication networks in the 3G stage, hoping to partially realize the automation and intelligence of the networks in the 3G stage." Dr. Ouyang Ye, one of the co-authors of the paper, Chief Technology Officer and Senior Vice President of AsiaInfo Technology recalled that in 2008, the industry began to define the self-organizing network at the international standard level for the first time.
It stands to reason that in the next 10 years, self-organizing networks should usher in a period of rising development, but this is not the case.
"In the past 10 years, American and some European telecom operators have deployed some self-optimizing network (SON) systems independently or in a mixed way, trying to integrate artificial intelligence into the planning, construction, maintenance and optimization of communication networks, but the overall effect has not reached expectations." Ouyang Ye said that this happened because the 2G and 3G networks themselves, including their ecology, software and hardware, equipment interfaces, etc., were not built according to the concept of intelligence, and the network elements themselves did not support artificial intelligence in the stage of standardization construction, and it was difficult to connect with artificial intelligence, network interfaces and signaling systems. It’s like inviting a famous doctor to treat a disease, but the patient himself doesn’t accept injections and medicines.
3G network is like an "iron plate", and the integration with artificial intelligence has become "dancing in chains".
It was not until the emergence of 5G that this situation ushered in a turning point.
Zhang Yaqin explained: "A very important feature of 5G is that at the protocol level, the Internet and communication networks have truly integrated. Therefore, many of our algorithms in the computer field and technologies in the Internet field can be more easily implemented in 5G networks. "
Therefore, in the construction of 5G infrastructure in China, we should consider the integration of artificial intelligence and leave an interface for the integration of artificial intelligence.
To this end, the author of this paper calls on the builders of the communication industry to further open up the network standardization interface, give artificial intelligence the opportunity to fully empower and inject wisdom into the network infrastructure, network management operation and maintenance system, business support system, etc., and take 5G and artificial intelligence as technology "combination" to further release the potential.
Wisdom "neurons" germinate in desk work
Back in February, 2017, the Second Working Group on Services and Systems of the International Organization for Standardization 3GPP began to study the intelligent network elements of the 5G core network. They defined a network element with network data analysis function: the nodes of a communication network can collect personalized information of users, form a unique personal portrait, analyze it through the portrait, and then provide the analysis results to the decision-making level.
This is the first time that mobile communication has defined standardization in the core network architecture since 1G to 5G, and requires the deployment of network artificial intelligence network elements.
"The intelligentization of 5G starts from standardization, which is a characteristic of the communication field, because the so-called communication must be connected to transmit information. And standardization is to avoid ‘ Chickens talk to ducks ’ The premise. " Ouyang Wei said.
On the other hand, the network element with "soul" is the foothold of artificial intelligence algorithm and model in 5G network. Through the network element, the ability of artificial intelligence is "injected" into the 5G network, and the nodes in the 5G network will gradually have different intelligent functions, and the network element will be able to participate in the decision control of the control plane of the core network in a comprehensive and real-time manner.
Since then, the new components of network intelligence have been enriched.
In June 2017, china communications standards association launched the research on the application of artificial intelligence in telecommunication networks.
In February 2018, the Open Wireless Access Network Alliance was established to open the "Android version" of the mobile information artificial intelligence framework.
In June, 2020, China Mobile and AsiaInfo Technology formally introduced the concept of federated learning into the 3GPR17 standard for the first time, forming the first global international standard of federated learning in the 5G field.
"In terms of 5G network intelligence, China operators and manufacturers are in the first camp." Ouyang Ye said that at present, in the key standards organizations for the intelligentization of 5G communication networks (including 3GPP, Global System for Mobile Communications, etc.), China manufacturers have a very high degree of participation, and together with more international manufacturers, they are working together to promote the intelligentization of 5G.
In 2023, the mobile communication network will realize initial intelligence.
"We expect that by 2023, the mobile communication network will realize initial network intelligence and initially build a network element of network artificial intelligence." Ouyang Ye told reporters that by this step, communication artificial intelligence will reach L2 level (primary intelligent autonomous network).
From 2023 to 2027, the communication network will evolve to L3 or even L4 level. At that time, the intelligent network can do more advanced things, such as "catching thieves".
Wei Leping gave an intelligent scene of transmission network. "The optical fiber is cut off, which happens many times every year. At present, an optical fiber bears the capacity of 8T, and once it is cut off, it is very harmful. But with artificial intelligence, you can make an early warning. " Wei Leping said that the optical fiber must be stressed before being cut off. Once stressed, sensitive artificial intelligence can know in advance that the signal is abnormal and even locate it accurately.
The attitude of relevant experts is relatively conservative as to whether the complete autonomy of the 5G intelligent network can be achieved by 2028.
"In 2027 and 2028, it is very likely that the first version of 6G international industry standard will be released. This key node determines whether artificial intelligence can continue to play a core role in the process of evolution to 6G. If the attempt at 5G fails, its importance and status will also become lower." Ouyang Wei said.
In Zhang Yaqin’s view, in the era of data 3.0 where everything is data, a network that can’t learn autonomy is embarrassing and useful.
Whether 5G or 6G communication networks can achieve a perfect high degree of autonomy in the next 10 years is unknown, but the dream of "unmanned" runs through the hearts of every scientific researcher and industrial worker who is obsessed with artificial intelligence.
关于作者