AI technology is rapidly converging in engineering simulation and design tools. It shows great potential for convergence, especially in the next-generation communications and electromagnetic fields, and experts from various fields are exploring directions and presenting outlooks.

MATLAB Expo 2024 Korea was held on the 11th at the Grand InterContinental Seoul Parnas in Samseong-dong, Seoul.

The event brought together numerous engineers and developers from across the country to participate in various technical conference sessions of MathWorks. There were sessions where we could see the flow of AI convergence vision, especially in the field of electromagnetic waves.

In system design, AI is expected to excel in future electromagnetic wave-related fields such as wireless communications, signal processing, and radar processing.

■ “Wireless Challenge, Solved with AI”
“We live in a hyper-connected world, with 90 percent of the world’s population owning a mobile phone,” said Arun Mulpur, MathWorks Industry Director, in a MathWorks keynote. “MATLAB is contributing to the advancement of 5G by leveraging its capabilities in 5G development and deployment.”

He noted that “MathWorks plans to continue investing in next-generation communications technologies and AI, and has upcoming releases that will enable MathWorks tools to model, simulate, and test these technologies.”

A major trend is that AI will be added to the 5G advanced standard, and many companies are using MathWorks tools for AI to develop wireless communications. Examples of use cases include △compressing channel state information from receiver to transmitter, △reducing the search space of beam pairs in MIMO systems, and △improving location accuracy.

Current challenges in wireless communications include: difficult modeling, computational complexity, efficient modem parameter optimization, and nonlinearity issues. Accordingly, the application of AI can provide an appropriate solution to difficult modeling in wireless, and it is expected that nonlinear function modeling will be possible.

■ Generative AI-based electromagnetic wave analysis
Park Hak-byeong, Senior Vice President of Samsung Electronics, gave a presentation in the special track titled ‘Electromagnetic Wave Analysis Using Generative AI: Application Example for Characteristic Evaluation of Array Antennas.’

We shared a case study on the characterization and analysis of array antennas using near-field, and introduced the process and results of antenna modeling and simulation, data preprocessing, and transfer learning using MATLAB using generative AI. Through this, we forecast the impact that generative AI will have on future electromagnetic wave design and analysis fields.

Senior Park said, “AI technology has been applied to electromagnetic wave design and analysis for several years now, and deep learning in particular is increasingly being used in manufacturing and design,” adding, “On the other hand, generative AI, which is used in various fields, does not yet have a clear direction in the field of electromagnetic wave technology.”

Since last year, Chief Park has been trying and analyzing automation of electromagnetic wave design and measurement from various perspectives through generative AI and LLM, and presented some of the results in this session today.

In order to verify the normal operation of the array antenna element, near-field measurement images can be learned through deep learning, and AI modeling can be used to determine whether the measurement value passes or fails. Senior Park compared this with experiments in which people directly modeled it and experiments in which generative AI-based modeling was conducted using Chat GPT and Claude 3.

Park said that the modeling experiment process was automated by using MATLAB simulation data instead of measurement data, and based on this learning data, deep learning modeling and code generation were performed, and explainable AI (XAI) was implemented using Grad-CAM.

As a result, Chief Park evaluated that the emergence of generative AI has lowered the hurdle of basic knowledge on necessary technologies, and made the observation that generative AI such as ChatGPT can replace the cooperation between fields that was necessary in the past. “As AI technologies gradually penetrate our surroundings, it is time to think about the work capabilities and methods that utilize them,” said Park.

■ AI Secretary, Become an Expert
Nvidia CEO Jensen Huang reiterated at Computex that the barriers to programming are lowering, and that anyone can code and become a programmer by simply talking to a computer. Generative AI is getting smarter and is now even invading the realm of experts.

The company's software stack generates code, automatically performs tests, and detects defects. It can also generate simulation data to drive the modeling process without having to create real-world data.

“Software is the language of automation,” said MathWorks director Arun Malpher, quoting CEO Jensen Huang. A software stack infused with AI accelerates this automation.

MathWorks first launched a service called MATLAB AI Chat Playground in 2023. With continuous feature improvements, it now generates initial MATLAB code drafts and is trained on data such as MATLAB white papers and examples to answer expert questions from users. Additionally, you can receive support for contents such as △deep learning △statistics and machine learning △optimization △toolboxes for control systems and signal processing through AI Playground.

AI assistants not only have the potential to expand into more specialized areas, but also break down barriers for those who lack basic knowledge by communicating and creating through natural language. As Chief Park Hak-byeong said, the hurdle to basic knowledge is lowering due to advanced AI technology, so it seems necessary to seek a future that prepares for this.