[Editor's Note] Artificial intelligence is converging across industries. With active attempts to apply AI in the electromagnetic field, we looked into related technologies, industries, and trends through the 'Electromagnetic Waves and AI' series that explores insights into the convergence of AI radio waves. MathWorks provides solutions across industries including finance, robots, industrial automation, automobiles, defense and aerospace, bio, machinery, and semiconductors, and is widely known for MATLAB and Simulink, which have over 130 tools. We met with MathWorks Korea Managing Director Kim Young-woo, who provides technology solutions across industries, to hear about the convergence of radio waves in the AI era and industrial trends.

■ You have been working at MathWorks for a long time. What is your experience at MathWorks?

I worked at Samsung Electronics for about 19 years and worked at Intel Korea R&D Center. I mainly worked on projects in the wireless communications field. Since then, I have been working at MathWorks Korea for 17 years.

Whereas my previous position was in the development team, this position is closer to the technology application service team. The appeal is that it allows me to experience various industries, outside of my previous experience in the wireless communications field, and directly identify customer problems and concerns.

For example, when a customer's research and development project is in progress, we guide them on what their complaints are and what areas we can improve during the design and mass production process, and we support them with consulting and training.

■ MathWorks platform is also evolving in line with the AI era. What makes MathWorks more essential in today’s industrial trends than in the past?

For example, in the automotive industry, the number of software lines required for a car has increased from 1 million lines in the 2000s to over 100 million lines today. It is expected that at least 500 million lines will be required to operate autonomous driving systems.

What this means is that many more lines of software need to be verified, and as the number of lines increases, the risk of human error increases proportionally. In particular, since automobiles are directly ridden by humans and provide convenience, system problems can easily lead to life-threatening situations, so functional safety is of the utmost importance.

Cutting-edge products such as drones, autonomous driving, and unmanned aerial vehicles are becoming more complex as electronics, machinery, and software are integrated, and AI functions that recognize the surroundings are added, making team collaboration difficult.

In this way, model-based design can lead to shortened development time, reduced costs, and timely release. In the existing electronic, mechanical, and software development, the design workflow documents the design requirements and distributes them to each team, and each developer in charge develops with each tool. In this case, the requirements described in the design specification may be ambiguous, and there may be cases where they are systematically impossible to realize.

These potential problems also mean that most errors in electromechanical software are discovered in the final integration testing phase. As the revision-redesign process is repeated, developers become burned out and companies suffer excessive consumption of human, material, and time resources.

As systems become more complex, the MathWorks platform is gradually expanding to accommodate the need for model-based design that minimizes redesign and modification.

■ Why is model-based design important?

Starting with the aerospace and automotive industries, standardized design and development workflows are becoming established toward model-based design, and this trend is leading to model-based design becoming widespread across industries.

If electronics, machinery, and software are integrated into a model rather than separate tools, system requirements can be verified through integrated system simulation that can be performed during the initial design. In addition, by utilizing the optimization function, it is possible to design the system design according to the system requirements by optimizing the variable elements of the system design. For this reason, many companies are currently seeking efficiency in product development through model-based design.

Another feature of model-based design is that once a model is completed, it can be reused in other related product designs. With just this model, you can generate C/C++ code and HDL (Hardware Description Language) VHDL, Verilog, and even CUDA using GPU coder through the automatic code generation function. For this reason, leading companies consider these models as important assets.

This automatic code generation is possible only when there is strong verification functionality, and traceability is secured from requirements to models and from models to automatic code generation.


In particular, communication between engineers in different fields of expertise such as electronics, machinery, and software can only become smoother through integrated model-based design.

■ There is talk that Chat GPT will find its place in electromagnetic design and analysis. What do you think about this?

I think it's a good opportunity. MATLAB is currently building and providing an AI Chat Playground based on OpenAI. It was first released around May 2023 and is continuously updated to improve its functions.

AI Chat Playground can generate MATLAB code drafts and answer users' questions.

I cautiously wonder if there will be a time in the future when MATLAB can be used through natural language. I expect the launch of AI Chat Playground to have the effect of lowering the barrier to entry for users.

There are over 130 MathWorks tools based on MATLAB. Some engineers may use them well, but if an engineer uses MATLAB intermittently, he or she may not remember how to use it or may have difficulty implementing functions. The emergence of Chat GPT lowers this barrier to entry.

MathWorks also offers a variety of training courses through professional trainers, and we believe that we can provide a more in-depth service by linking these training courses.

Personally, I look forward to a time when we can issue commands to Chat GPT with just a concept and create algorithms and programs.

■ How is AI convergence progressing in the electromagnetic field?

The biggest hurdle for wireless systems is that they are invisible. They need to be visible to perform supervised learning, but the only way to train an invisible wireless system is through a simulator.

Data is received through a simulator or digital twin, a virtual AI model is created, and this is verified to see how well it operates in real life. These AI models can be linked with Simulink and MATLAB to enable wireless system-level simulation.


If we were to create a drone, we would first create a virtual scene with a drone equipped with virtual sensors. The drone has virtual sensors such as cameras, radars, and lidars in operation. By placing an AI model on top of this system, we can check in advance in a virtual space how well it is working. I think this is the advantage of our platform.

RF and radar can also apply AI models. When there is an AI model, it does not end with creating it, but must be verified on the system. MathWorks products are released with various examples to meet customer needs, and various AI solutions are provided using image data extracted from the simulator.


■ A final word to e4ds news readers

Systems are becoming increasingly complex. As the needs of the 4th industrial revolution grow and digital engineering gains attention, I believe it is time to find new ways to solve problems and approach them.

The gap between those with and without infrastructure will only grow wider in complex systems, such as those that improve productivity, reduce human errors, and improve verification efficiency through simulations in virtual environments. AI is just one system element. When linking it with a system, keep in mind the gap between those with and without infrastructure.

thank you