Dynamic spectrum management technology: from cognitive radio to artificial intelligence

With the rapid development of radio technology, limited frequency resources have been unable to meet the increasing demand for frequency. In order to improve the utilization efficiency of frequency resources, comprehensive management of frequency resources has received extensive attention. In the process of traditional static management to dynamic management, new technologies such as cognitive radio and TDOA positioning have developed rapidly. However, the most cutting-edge artificial intelligence technology will complete data-to-decision conversion more efficiently and complete comprehensive tasks more intelligently.

Cognitive Radio

Solve spectrum utilization problems

Dynamic spectrum sharing is an important topic in the research of cognitive wireless networks, aiming to improve the utilization of wireless spectrum.

At present, wireless networks generally adopt a fixed spectrum allocation method, and almost all wireless terminals work under a spectrum allocated by some spectrum management organizations (such as the International Telecommunication Union and national spectrum management agencies). Studies have shown that in this mode, most of the allocated spectrum is often not fully used in many areas, and its utilization rate ranges from 15% to 85%. Dynamic spectrum sharing is an important topic in cognitive wireless network research, aiming to improve the utilization efficiency of wireless spectrum resources. Driven by opportunities, cognitive radios allow wireless terminals to automatically perceive, identify, and utilize any idle spectrum resources. Once an authorized user on the used spectrum segment appears, the wireless terminal will actively give up the corresponding spectrum and switch to another available spectrum.

In order to solve the problem of spectrum dynamic utilization, cognitive radio technology has developed rapidly in recent years. The research of cognitive radio mainly includes: the next generation wireless communication (xG) project funded by the Defense Advanced Research Projects Agency (DARPA), and the National Natural Science Foundation of China, which is doing a cognitive wireless technology project at the Winlab Laboratory of Rutgers University. Research on adaptive radio frequency technology in the UK's mobile telecommunications technology virtual center, DRIVE, OverDRiVE and TRUST funded by the European Telecommunications Association, cooperation and cross-layer design techniques, spatial signal detection and analysis and QoS in the national "863" program cognitive radio system Guarantee mechanism, etc.

Cognitive radio is a radio technology that can change the parameters of the transmitting end according to environmental changes. It can use this frequency band when no user uses the licensed frequency band, which greatly improves the spectrum utilization, makes up for the defects of fixed spectrum allocation, and is the key technology of the next generation network. From a technical point of view, cognitive radios can change transmitter parameters based on environmental changes. It is used for adaptive spectrum management and subsystem development - smart antennas, sensors and receivers, adaptive modulation and waveform technology, etc., is the key to the dynamic use of spectrum in next generation networks. Cognitive radio can use this band when no user uses the licensed band, greatly improving spectrum utilization and making up for the shortcomings of fixed spectrum allocation.

Cognitive radio has two main capabilities. One is cognitive ability. Cognitive ability is the ability to acquire perceptual information from the environment. Use complex techniques to obtain instantaneous spatial variables in the environment and avoid interference with other users. The other is the ability to reset. Cognitive capabilities perceive the spectrum, while resetting capabilities allow the radio to dynamically configure hardware parameters to transmit and receive on different frequencies, as well as different transport accesses supported by hardware devices. Therefore, cognitive radio can perform multiple functions in dynamic spectrum management. The first is spectrum sensing, which spectrum is available and detects if an authorized user exists when the user is working on an authorized frequency band. This is followed by spectrum management, which selects the best available channel usage. Again, spectrum sharing, adjusting channel access with other users, provides users with an appropriate spectrum arrangement. Finally, the spectrum is moved, and after detecting the time-space channel of the authorized user, it migrates to other frequency bands.

The shortcoming of cognitive radio technology is that it greatly increases the complexity of the system, and the communication quality cannot be completely guaranteed. Therefore, a new system, namely the dynamic authorization management system, has been proposed internationally. In other words, companies such as broadcasting and television are allowed to lease unused spectrum to users in hotspots. The main representatives of the international research frontier are the LSA of the European Union and the SuperWiFi of the United States. Although cognitive radio partially solves the dynamic spectrum management problem of a certain frequency band from the perspective of the radio user, there is still no effective solution for the dynamic management of the entire spectrum by the radio regulatory agency. In the technical analysis of radio spectrum management, it is often necessary to identify the key signals as well as the anomalous signals. In the case of multiple signal types and multiple transmitting stations in the frequency band, multiple signals will overlap in the same frequency band, posing a huge challenge to spectrum management. Spectrum sensing in cognitive radio technology can quickly identify the idle spectrum and establish a pool of free spectrum, but the dynamic allocation of spectrum cannot be solved for the changing radio spectrum. The dynamic management of spectrum needs to be solved not only by the dynamic sensing spectrum environment, but also by the close decision of self-determination to achieve dynamic management. The rise of artificial intelligence may bring development opportunities for dynamic spectrum management.

Artificial intelligence commercialization

Rapid development

With the development of technology, artificial intelligence has been applied in e-commerce, finance and medical.

The term artificial intelligence was first proposed by cognitive scientist John McCarthy in his research. His interpretation of artificial intelligence is a speculation in this study that any learning behavior or other intellectual characteristics can be accurately described in principle, so that it can be manufactured. Take out a machine to simulate it. With the development of technology, artificial intelligence has been applied in e-commerce, finance and medical. At the same time, machine learning and deep learning are often mentioned together with artificial intelligence. Machine learning is a way or subset of artificial intelligence that emphasizes learning rather than computer programs. A machine uses sophisticated algorithms to analyze large amounts of data, identify patterns in the data, and make predictions. Deep learning is a subset of machine learning that uses large amounts of data and computational power to simulate deep neural networks. These three concepts are closely related, but each has its own focus.

Artificial intelligence was officially presented at the Dartmouth Conference in 1956, and entered a period of rapid development in 2006. With the breakthrough of deep learning algorithms in speech and image recognition, the commercialization of artificial intelligence has achieved rapid development. In 2016, AlphaGo defeated Li Shishi, and artificial intelligence received unprecedented attention from the world. Artificial intelligence products and services continue to be introduced, such as Amazon Echo smart speakers, Facebook to use artificial intelligence to enhance user experience, have been widely recognized in the market, BAT is also actively promoting artificial intelligence projects.

In terms of policy support, the government has vigorously supported the artificial intelligence industry. In the "New Generation Artificial Intelligence Development Plan" released in July this year, it is proposed that by 2020, the overall technology and application of artificial intelligence in China will be synchronized with the world advanced level; by 2025, China The basic theory of artificial intelligence has achieved a major breakthrough, and some technologies and applications have reached the world's leading level. By 2030, China's artificial intelligence theory, technology and application will reach the world's leading level and become the world's major artificial intelligence innovation center.

The “Artificial Intelligence for Humanity Global Summit”, held in Geneva from June 7th to 9th, 2017, aims to accelerate the development and popularization of artificial intelligence (AI) solutions to address poverty, hunger, health, education, equality and environmental protection. Waiting for global challenges. As the UN's specialized agency for ICT, ITU aims to guide artificial intelligence to continuously achieve innovation to achieve the goal of sustainable development of the United Nations. ITU Secretary-General Houlin Zhao said: "We are providing a neutral for international dialogue. Platform to reach a consensus on the capabilities of emerging artificial intelligence technologies."

“In many of our public competitions, we can see that each team uses artificial intelligence as a basic tool in many fields, from creating personalized learning experiences for Tanzanian children who cannot get formal education, to empowering consumers to use medical three-record instruments. The device makes medical decisions, then guides advanced, autonomous robotic vehicles to explore the deep ocean or find a path on the moon's surface.” XPRIZE CEO (CEO) MarcusShingles said, “We recognize that with With the accelerated advancement and popularization of artificial intelligence, the new generation of problem solvers face great opportunities in meeting global challenges.” This event is the first meeting of the annual series of artificial intelligence, from government, industry, UN agencies. Representatives from various groups in the civil society and artificial intelligence research community explored the latest developments in artificial intelligence and its impact on issues such as regulation, ethics, and security and privacy.

Artificial intelligence on

New spectrum management model

In order to solve the problem of shortage of electromagnetic spectrum resources, artificial intelligence will focus on intelligent decision-making.

Professor Wu Qihui, a spectrum research expert, said at the “2017 Global Future Network Development Summit” that traditional spectrum decision-making is a manual method, mainly because the situation is relatively simple, and may not require decision-making, or even just predict it. But now spectrum operations are carried out in a complex electromagnetic spectrum environment, the complexity is mainly reflected in diversity, intensive, large-scale, high dynamics and high confrontation. We study intelligent spectrum decision-making or autonomous spectrum decision-making. From the operational point of view, we mainly solve pre-war rapid planning, wartime self-coordination and confrontation with the enemy. It mainly uses the intelligent decision-making method of human-machine hybrid to make advance decision and make temporary decision.

Driven by the mobile Internet, the Internet of Things, and the integrated information network, the future wireless networks will develop in the direction of higher speed, more access, and wider coverage, posing more challenges to spectrum resources. In order to cope with these three challenges, we need to make three changes in the spectrum. These three changes also reflect Internet +, artificial intelligence +, and spectrum transformation.

In order to solve the problem of shortage of electromagnetic spectrum resources and promote the transformation of spectrum resources from static monopoly to dynamic sharing, artificial intelligence will focus on intelligent decision-making, promoting the transition from isolated monitoring to grid monitoring and analysis, while artificially in complex electromagnetic environment. Decision-making changes to autonomous decision-making. The radio spectrum machine learning system is a technical application of artificial intelligence in radio frequency management.

The Radio Spectrum Machine Learning System funded by the Defense Advanced Research Projects Agency (DARPA) consists of four major technical components:

1. Feature learning: Identify signals from signal data and classify them according to user settings.

2. Intelligent monitoring: intelligently pay attention to the key frequency bands or frequency points in the spectrum from the massive data collected in real time. Predict and adjust to the corresponding key monitoring frequency band or frequency point according to the rules set by the user.

3. Automatic Perceptual Recognition: Automatically adjusts monitoring settings based on user task needs.

4. Signal Synthesis: Digitally synthesize signals based on user needs and improve synthesized signal quality.

In the technical analysis process of radio spectrum management, it is often necessary to identify key signals and abnormal signals. This usually depends on the experience of monitoring facilities and engineers. Once encountering unexpected situations such as black broadcasting and pseudo base stations, it is often necessary to invest a lot of human resources. Time inspection positioning. In addition, in order to improve the efficiency of frequency use, the management department hopes to improve the frequency band sharing technology and predict the use of the frequency band in order to perform frequency reuse without causing interference.

Admittedly, the application of artificial intelligence in radio management faces many challenges. For example, the application of artificial intelligence in dynamic spectrum management is big data. The data required is not only large but also complex. Radio monitoring data, frequency data, and station data are all focused but inseparable. On the other hand, deep learning requires a rigorous set of rules for self-determination and requires a quantifiable assessment of the spectrum. The National Radio Management Plan (2016-2020) states that during the 13th Five-Year Plan period, the primary task is to innovate spectrum management, establish a scientific and rational spectrum utilization assessment and frequency recovery mechanism, and form an administrative approval and market-oriented configuration management system. Therefore, on the one hand, we will lay a solid foundation, on the other hand, we must keep up with forward-looking technology development trends, use artificial intelligence technology to serve spectrum dynamic management, and serve radio management under the new situation.