Home  >  Article  >  Technology peripherals  >  As wireless system complexity increases, AI becomes the key to overcoming challenges

As wireless system complexity increases, AI becomes the key to overcoming challenges

PHPz
PHPzforward
2023-04-12 11:10:131320browse

As mobile wireless technology makes the leap to 5G, the complexity of wireless system design is increasing.

Currently, due to the increasing demand for expanding user groups, it is necessary to strengthen the optimization and sharing of precious resources, and it also increases the difficulty of wireless network management. These adjustments are forcing engineers to break through the traditional rule-based methods and find new solutions. AI becomes their go-to solution to modern system challenges.

Recently, Houman Zarrinkoub, principal product manager of MathWorks, pointed out in the article "The key to overcoming complexity in modern wireless systems design" that from managing communications between autonomous vehicles to optimizing mobile calls Resource Allocation,AI brings necessary complexity to the development of,modern wireless applications.

Today, as the number and range of devices connected to the network expands, the importance of AI in the wireless field has greatly increased. Engineers must be prepared to introduce AI into increasingly complex systems and understand the benefits and applications of AI in wireless systems, as well as best practices in implementation, which will be key to the future success of wireless system technology, Houman Zarrinkoub said.

1 Benefits of AI to Wireless Systems

The transition to 5G brings optimization of mobile broadband network speed and quality, and The need for ultra-reliable low-rate and large-scale machine communications for time-sensitive connections between Industry 4.0 devices – these are three distinct use cases in modern networks and the competitiveness driving engineers to adopt AI.

As devices compete for network resources and the number of users and applications in wireless systems continues to increase, linear design patterns once understood as human-based rules fall short. However, by automatically and efficiently extracting arbitrary patterns, AI can better solve nonlinear problems beyond the capabilities of human-based methods.

In this context, artificial intelligence refers to those machine learning and deep learning systems used to identify patterns in connected devices, people’s communication channels, and the resources given to that link. Make optimizations to improve performance. Simply put, running a network for these different use cases without leveraging AI methods is an almost impossible task.

In addition, artificial intelligence is also helpful in project management. By estimating the behavior of the source environment, integrating the simulation environment into the algorithm model allows engineers to study the main effects of the system more quickly with minimal computing resources, leaving more time for exploring the design and subsequent iterations, reducing costs. and development time.

As wireless system complexity increases, AI becomes the key to overcoming challenges

Note: Workflow of AI for Wireless - data generation, AI training, verification and testing, and deployment on hardware

2 Best practices for applying AI in wireless systems

Entering the application stage, data size and quality are critical to the effectiveness of the AI ​​model Deployment plays a crucial role.

To handle a range of real-world scenarios, these models need to be trained using a wide range of data. By synthesizing new data based on primitives, or extracting them from wireless signals, wireless system applications will also provide 5G network designers with the changes in data needed to robustly train AI. Without a large training data set on which to iterate on different algorithms, the end result may be a narrow local optimization rather than a global optimization of the whole.

Additionally, a robust approach to testing AI models in the field is critical.

One of the issues with signal variation required to test AI technologies is that signals captured in narrow, localized geographic environments can adversely affect how engineers optimize design quality. Without field iteration, parameters for individual cases will also not be used to optimize the AI ​​for specific locations, negatively impacting call performance.

3 Main application areas of AI in the wireless world

Digital transformation in fields such as telecommunications and automobiles also requires the participation of AI, as does AI The main driver of these applications.

As applications such as smart cities, telecommunications networks and autonomous vehicles (AVs) become more connected, electronic communications are capable of generating vast amounts of data when placed in areas that were once machine-oriented. Added network resources will also become stretched thin.

In the telecommunications field, artificial intelligence is deployed at two levels - the physical layer (PHY) and above the PHY. Among them, the AI ​​application used to improve the performance of connecting two user lines is called For operation in the PHY. The application of AI technology at the physical layer includes digital predistortion, channel estimation and channel resource optimization, as well as automatically adjusting transceiver parameters during a call, which can also be called autoencoder design.

Channel optimization refers to enhancing the connection between two devices, especially the connection between network infrastructure and user equipment. Often, this also means using AI to overcome signal variability in the local environment through techniques such as fingerprinting and channel state information compression.

Through fingerprint recognition, AI can map interference to propagation patterns in indoor environments (caused by personal entry) to optimize the positioning of wireless networks. AI will personalize 5G signals based on these changes to estimate the user's location. At the same time, channel state information compression can use AI to compress the feedback data from the user equipment to the base station, ensuring that the feedback loop that informs the base station to try to improve call performance does not exceed the available bandwidth, causing call interruption.

Above-PHY is mainly used for network management and resource allocation, such as scheduling, beam management and spectrum allocation. It refers to the function of managing and optimizing core system resources and can be used in the network Competing users and use cases. As the number of network users and use cases increases, network designers have turned to artificial intelligence technology to respond to distribution needs in real time.

In the automotive field, wireless connectivity using AI makes safe autonomous driving possible. Autonomous vehicles (AVs) rely on data from multiple sources, including lidar, radar, and wireless sensors, to interpret their environment. The hardware in self-driving cars needs to process data from many competing signals, and AI can achieve sensor fusion to fuse the competing signals so that the vehicle software can understand its location and determine how it interacts with the environment.

As use cases for wireless technology expand, so does the need to apply artificial intelligence in these systems. Without AI, systems such as 5G, self-driving cars, and IoT applications will not have the complexity required to operate effectively. While the role of AI in engineering, and specifically in wireless system design, has been increasing in recent years, it can be expected to continue to rise at an even faster rate as the number of use cases and network users grows.

The above is the detailed content of As wireless system complexity increases, AI becomes the key to overcoming challenges. For more information, please follow other related articles on the PHP Chinese website!

Statement:
This article is reproduced at:51cto.com. If there is any infringement, please contact admin@php.cn delete