Wireless communication plays an important role in our daily lives. The latest 5G network communication aims to provide faster, more reliable, and more efficient data transmission speeds and services than the existing 4G networks. Meanwhile, the rapid development of artificial intelligence in recent years has provided new methods for solving problems in traditional communication technologies, opening up new opportunities. This article combines the research progress of a classic technology in the field of artificial intelligence—machine learning—and introduces its applications in advanced wireless communication of 5G, further exploring the challenges faced by machine learning in 5G advanced wireless communication and future research directions.
1. Introduction to 5G Network Communication
5G networks, the fifth generation of mobile communication technology, are the core of the new generation of mobile communication technology. 5G has ultra-high spectrum utilization and energy efficiency, improving transmission rates and resource utilization by at least one order of magnitude or more compared to 4G mobile communication, with significant improvements in wireless coverage performance, transmission delay, and system security. 5G mobile communication is closely integrated with other wireless mobile communication technologies, forming a new generation of ubiquitous mobile information networks. It can meet the development demand for a 1000-fold increase in mobile internet traffic over the next decade [1]. Furthermore, the application fields of 5G mobile communication have further expanded, bringing unprecedented convenience to people’s lives in areas such as smart transportation, remote healthcare, the Internet of Things, and virtual reality entertainment.
However, 5G communication still faces many challenges, such as the high speed and large capacity of the 5G network, which places higher demands on network equipment and information transmission technologies; while 5G networks are characterized by flexibility and adaptability, their security and stability require more technological innovation to ensure.
2. Introduction to Machine Learning Technology
With the development of artificial intelligence technologies, new ideas have been provided for problems that traditional methods cannot solve. As a major branch of artificial intelligence technology, machine learning solves complex problems by learning from previous data and extracting features.
Machine learning is mainly divided into three branches: supervised learning, unsupervised learning, and reinforcement learning. Figure 1 illustrates the relationship between Artificial Intelligence (AI), Machine Learning (ML), supervised learning, unsupervised learning, and reinforcement learning.
Supervised learning learns the corresponding mapping model by using a labeled training dataset. Deep Neural Networks (DNN) are a typical representative of supervised learning, obtaining the convergence coefficients of multilayer artificial neural networks through offline training and inferring new data.
In contrast to supervised learning, unsupervised learning uses an unlabeled training dataset to obtain mapping models. In a classic autoencoder structure, it adjusts the network coefficients to recover the true values of the input data by learning the inherent features of the input data.
Reinforcement learning is based on the dynamic interaction between an agent and the external environment, allowing for online data processing. The agent learns by exploring the environment and receiving rewards or penalties for its actions, optimizing cumulative rewards through the best sequence of actions.
Figure 1: Relationship between Artificial Intelligence, Machine Learning, Supervised Learning, Unsupervised Learning, and Reinforcement Learning
3. Applications of Machine Learning in 5G Communication
In recent years, global researchers have shown great interest in using machine learning to develop 5G communication technology. The following focuses on several applications of machine learning in 5G wireless communication.
3.1 Adaptive Modulation Coding Technology (AMC)
AMC is an adaptive coding modulation technology used on wireless channels, ensuring the transmission quality of the link by adjusting the modulation method and coding rate of the wireless link transmission.
However, current AMC technology, in practical applications, has not shown good performance due to inaccurate model approximations or overly large lookup table sizes that complicate the system. Since AMC is a typical classification problem, supervised learning in machine learning naturally becomes an important choice for optimizing adaptive modulation coding technology. The most commonly used supervised learning algorithm to solve this problem is the K-NN (K Nearest Neighbors) algorithm, which determines the category of a new value X based on the types of the K nearest points to it. For example, in Figure 2, when judging the type for k=3, we observe that among the three adjacent points, two are triangles and one is a square, so the type for k=3 is determined to be a triangle.
Figure 2: K-NN Algorithm Illustration
However, the biggest difficulty of the K-NN algorithm is that it cannot clearly define class boundaries [2], requiring a large offline training database. In contrast, reinforcement learning can learn directly from the environment. Therefore, researchers have attempted to use reinforcement learning based on the Markov process to implement AMC, which has higher adaptability compared to the K-NN algorithm.
3.2 Channel Equalization Technology
Channel equalization is a measure taken by communication systems to improve the transmission performance of fading channels, mainly to eliminate inter-symbol interference and nonlinear distortion problems. Its principle is to compensate based on the characteristics of the channel or the entire transmission system; for example, Figure 3 shows the channel equalization technology that adjusts transmission power based on channel quality.
Since machine learning can achieve adaptive signal processing capabilities, researchers have proposed machine learning-based equalizers to improve the characteristics of traditional equalizers. By extracting key features of time-varying wireless channels through machine learning, an adaptive equalizer based on Multilayer Perceptron (MLP) has been proposed, mainly used for suppressing inter-symbol interference (ISI) in linear channels. Subsequently, researchers proposed a recurrent neural network equalizer based on FLANN on the basis of the MLP equalizer, which simulates the capability of nonlinear filtering, used to equalize complex signals with nonlinear channel distortion, and has lower computational complexity.
Figure 3: Channel Equalization Technology Adjusting Transmission Power Based on Channel Quality
3.3 Load Prediction
Since the frequency of 5G wireless networks is much faster than that of 4G, and the higher the frequency, the faster the fading, more 5G base stations need to be deployed to cover the same area as 4G. Therefore, load prediction to overcome the energy consumption caused by base station density has become a key issue.
For many years, machine learning has been widely used in load prediction. Various machine learning algorithms can be utilized. The Autoregressive Integrated Moving Average (ARIMA) model is a simple load prediction method that has been widely applied; the Prophet model achieves predictions through time series analysis and can also be used for load prediction; Long Short-Term Memory (LSTM) is a multi-memory unit structure based on RNN, suitable for load prediction. Ensemble Learning (EL) can combine multiple models linearly to generate an ensemble model with strong predictive capabilities. Each ML algorithm has its own advantages and disadvantages. Therefore, the most suitable machine learning algorithm for load prediction needs to be determined based on different application goals.
4. Challenges and Future Directions
Although machine learning is now widely applied in various technologies of 5G communication, there are still some problems and challenges, such as:
1. Current machine learning methods require long convergence times, which limits their application in dynamic wireless communication; therefore, further research is needed to accelerate the convergence process of machine learning.
2. Due to the time-varying characteristics of wireless channels, machine learning parameters and even methods need to be continuously adjusted, greatly increasing the complexity of the communication system. Therefore, in-depth research is needed on unified machine learning methods for different applications.
3. Most current machine learning methods require a large amount of labeled data for model learning and training, but most of the data generated by wireless networks are unlabeled raw data. Therefore, further research and improvement on the labeling of raw data are needed.
References:
[1] You Xiaohu, Pan Zhiwen, Gao Xiqi, et al. Development Trends and Key Technologies of 5G Mobile Communication [J]. China Science: Information Science, 2014, 44(5):551-563. DOI:10.1360/N112014-00032.
[2] Y. Zhou, J. Chen, M. Zhang, D. Li and Y. Gao, “Applications of Machine Learning for 5G Advanced Wireless Systems,” 2021 International Wireless Communications and Mobile Computing (IWCMC), Harbin City, China, 2021, pp. 1700-1704, doi: 10.1109/IWCMC51323.2021.9498754.
Source: China Mobile Association