Richard Rovner, VP of MathWorks (Natick, Massachusetts), explained the importance of MATLAB EXPO in communicating with users, “With over 130 software products and releases twice a year, we know our users have a really hard time keeping current with all the new capabilities. We have lots of information on our website and regular digital communications, but you are dedicated and focused when you come here for a day. The event is aimed at our users, and it’s an opportunity for them to learn about what’s new, the software they more than likely have access to but have yet to use, and to learn from each other.”

 

Modeling and simulating embedded systems

 

To many embedded system developers, using modeling and simulation tools for straightforward use cases might appear unnecessary. However, they are the norm for designing highly complicated systems, particularly for those operating in multi-domain environments, such as those for aerospace flight controls and autonomous vehicles. EE Times Europe asked Rovner whether modeling and simulation are more important as embedded systems become more complex. “Yes, over the last two decades, we’ve seen growth in the complexity of these embedded systems, and it’s happening in more and more applications. Everyone knows about aero and auto, but take industrial automation, a sector where complex system use is rising. Overall, we see tremendous growth in complex systems, and this stems from incorporating software into more of the design process from the beginning. Software design is becoming more ingrained in the whole workflow.”

 

To illustrate the increasing complexity of embedded systems and the reasons to model and simulate a design, Rovner gave an example of an intelligent sensor application, “You might want to include AI, but that involves selecting the correct algorithm, planning the training and testing of the algorithm, then you have to think about the embedded design, and deploying it to the network—also, there are trends such as digital twins or autonomous operation to consider. Once in the field, the sensors need monitoring and updating. Even for small, straightforward devices, there are opportunities to think about it from a model-based design perspective. That also gives you a platform to work with, and once you have that platform, you will build similar devices much more quickly. We have a lot of users in the major industrial corporations, but we also have thousands of startup companies using MATLAB and Simulink from the beginning.”

 

Inception to deployment as quickly as possible

 

Jos Martin, director of engineering at MathWorks, added another perspective regarding increasing system complexity, “What we thought of as simple 20 years ago is so complex compared to 20 years ago that what we now consider simple is still really hard.” Martin highlighted that what might have had 30,000 lines of code back then may have two million lines of code today. He then explained why startups are adopting MathWorks products from scratch, “Lots of startups are using MATLAB. I highlight them more because they aren’t doing anything overly complex. After all, they would need more people, but they are still developing something difficult, so they must do something quickly, cheaply, efficiently and complex enough to become a valuable product. Many startups use engineered systems, and they use our products because they want to go from inception of design to running the system as quickly as possible.”

 

EE Times Europe asked Martin whether there had been a fundamental way embedded developers approach system design. Did they have to change through necessity, or is there a new school of developers looking to do things quicker and smarter? He said, “Part of the change is through necessity. Twenty or so years ago, embedded systems focused on devices like MPC555. What could you do with it compared to today’s Arm Cortex device with a GPU? Today, we’re developing embedded systems with power budgets [the same as twenty years ago] that are radically different. It would be utterly trivial to do what we did 25 years ago on the hardware we now have. The market has pushed people to do clever things, and the demands of what the market expects you to be able to do with embedded hardware are such that complexity is just intrinsic because there’s so much more compute available.”

 

Increasing productivity

 

Martin highlighted the need for large and small organizations to be productive. He gave an example of a one-person development organization and how the person realized he would be far more productive using a system-level design tool, stating, “Productivity should always take the lead. I hope any embedded developer would have a system model in mind with an understanding of how a development fits together. That’s the system model approach. Use toolchains to make yourself effective because you’re not productive writing code. Writing code is quite hard work. You have to think about incredibly low-level things. By pushing the design aspects to higher-level concepts, you will likely be more productive and produce a higher-quality design.”

 

Engineered product? Think modeling and simulation

 

EE Times Europe asked Rovner what advice he would give to engineers and developers to encourage the adoption of modeling and simulation tools. “If you are at the point where you are about to design an engineered product that will have some level of performance, is intelligent and has a predictive capability, you need to think about modeling and simulation. Using this approach, you’ll get from prototype to product faster, cheaply and with fewer errors. You’ll also discover bugs early in the process rather than later when they become expensive.”