6 Trends in Embedded Systems Development That Will Affect Systems Engineering

6 Trends in Embedded Systems Development That Will Affect Systems Engineering

There is no question that a transformation is taking place in product development: New technologies are making new products possible, and are changing the way how we think about existing products, like cars. But what are the trends driving this transformation? This article outlines six trends, based on the keynote of Jean-Marc Chery (STMicroelectronics) at Embedded World 2019.

It’s hardly necessary to point out the changes happening right now, but Jean-Marc put them in Numbers. In five years (2017 – 2021) we can expect:

  • Number of wearable devices: 300 → 800 million
  • Number of networked smart home devices: 0.4 → 1.8 billion
  • Number of industrial devices, including retail and advertising: 4 → 10 billion
  • Number of smart driving devices: 1.1 → 2.2 billion

Theses numbers – and the possibilities the market offers – are driving six trends:

1. More Computing Power, New Architectures

Now that we know what is possible, we need the power (and memory) to realize it. And without it, we can’t get better performance, new features, new functions, reliability and safety.

But especially the last two (reliability and safety) will indirectly trigger another trend: We will see new scalable architectures that support these needs. For example, consider a modern car with up to 100 control units. Updating software and firmware is a nightmare. Consider a car with built-in GPS. The driver may still elect to use the smartphone for navigation, as updating the GPS would be too much hassle (trip to the garage, etc.). Over-the-air updates are hotly discussed, but security is often a challenge. So we see the related trend that fragmented architectures are aggregated.

2. Lower Power Consumption

Battery technology is catching up, but the best option is still to avoid consuming power in the first place. For this to work, we have to improve power consumption at every level: the (hardware) technology, during integration, in the software and in the system. Starting with production (silicon) technology. At some point, we have to decide what should be done in hardware for performance, and what in software for flexibility. Take the processing of sensor data as an example: The preprocessing can take place in hardware. The processed data can then be either sent to the cloud, or processed locally if needed.

3. More Security

Surprise! Embedded security is an issue, who would have thought? We already reported on this last year in  IoT Security: A Tsunami is coming.

More specifically, there are four areas that need to be addressed, which are:

  • Authenticity verification
  • Secure data exchanges
  • Secure data storage
  • Secure code execution

4. More Robustness

Robustness primarily applies to functional safety, which of course also applies on the hardware level. However, we this we understand well and have practiced for over a hundred years. Today, we must take software into account. Regulations are updated to reflect this. Again, this is applicable on many levels

5. Connectivity

When thinking about connectivity, we think about phones, home appliances and the like. But Connectivity has a much broader impact.For instance, robots employ short-range wireless to replace cables, as this is more reliable.Connectivity includes alignment of protocols and also relies on an infrastructure (network) to support it.

6. More local autonomy

Machines will take more and more their environment into account and make autonomous decisions. For embedded a resulting question is to where to do the processing: On the node? In the cloud? Or on the edge of the network? This requires a complex interplay of localized sensing and intelligence, distributed systems, edge computing and, yes, artificial intelligence. Getting this right takes a complex process: We need to classify the level of intelligence to understand the processor needs. Then, we train the neural network (independent of platform), and eventually convert the trained neural network into optimized code.

Bottom Line

These are exciting times for consumers and users of technology, but there are many challenges for getting this right. What we described here is just what we are seeing on the embedded front. But this is just one of the many things to consider in systems engineering.

Photo by Franck V. on Unsplash