Once again, I am delighted to share a summary of the embedded industry and a forecast of industry trends for this time around, 2024. How did last year’s predictions turn out? If you are curious, you can read the post from a year ago: Embedded systems in 2023 – forecasts and trends
Firstly, let’s tackle a topic that has impacted plenty (if not all!) of developers, namely AI.
Putting aside the debate over whether AI will replace developers, let’s delve deeper into the technology itself for now. Giants such as NVIDIA, Qualcomm, and OpenAI are at the forefront of investing in and developing the technology. Evidence of its downright frenetic growth in 2023 can be observed in NVIDIA’s Jetson Orin series of platforms. Among these, the most potent member of this family has a computational power of 275 TOPS and a claimed power consumption of up to 40W. We can process images up to 8K and 30FPS using this technology. If you need an embedded solution with lower power consumption, for instance, up to 15W, then the Jetson Orin Nano with a computing power of 40 TOPS should be ideal.
A question may arise regarding the difference in development between the Jetson Orin Nano and the Jetson Nano. Both platforms serve the same purpose. In the meanderings of the Internet, one can find results that are 5.4 times better in favor of the new platform for floating-point calculations used in AI and 80 times better for fixed-point calculations of up to 8 bits, which at first glance may seem negligible.
Another tidbit of information is the effective use of AI during embedded systems development. In Jacob Benningo’s articles: Unveiling the ultimate 5 AI tools for embedded software developers and 3 epic ways AI and ChatGPT are transforming embedded, one can find worthwhile and valuable hints on how to apply AI to optimize work on integrated projects. The author sees substantial potential in the use of AI assistants in elements such as suggestions during debugging and development, assistance with repetitive or boring coding elements, or searching for information in technical documentation. The symbiosis seems very promising and could provide embedded developers with a substantial boost in their efficiency.
This year, the automotive industry has passed under the banner of R155 and ISO 21434 regulations, and the imminent deadline for implementing CSMS, also known as the Cyber Security Management System. It is clear that vehicle manufacturers and their subcontractors have a lot on their plates, with ISO 21434’s update already announced. My participation in the adaptation of the standard has also made me notice that the organization’s maturity in terms of cyber security processes is not being taken into account, which is detrimental to the quality and speed of implementing changes.
IoT devices continue to be in the spotlight, both for hackers and enterprise executives. To aid this area of the industry, NIST 2019 launched a competition for the most sophisticated family of encryption algorithms for small embedded devices such as IoT (for so-called lightweight cryptography). In 2023, a winner has been selected – the proposed standard will be based on a family of algorithms called Ascon. This topic is covered in a more comprehensive paper for those interested.
There’s one thing I can say about the semiconductor market – we’re slowly moving in the right direction. Particularly when it comes to availability, since prices inexorably rise.
Some components are available off-the-shelf, though you may still have to queue in certain cases. Silicon manufacturers hint at further increases in 2024, arguing that these will be due to rising production costs. The main thread in the background is still diversification of production. Many places are being considered as new hubs of semiconductor development and production, with India leading the pack.
It is marvelous, as seen from the repositories and news stories. As sponsors and contributors join, the portfolio of supported platforms is growing (version 3.5 has been increased by 45), the number of drivers and ready-to-use solutions is soaring, and believe me, it is not easy to land a solution on the main branch. The code that ends up there is subject to relentless restrictions.
In addition, the latest version now integrates directly with the free static code analysis tool, CodeChecker. A robust native simulator, dynamic modules, and many other features have been added, which you can learn more about here.
In my opinion, Zephyr is coming to the forefront of its field and is slowly becoming a standard. This is like Linux on platforms with more computing capability.
Rust has continued to fight bravely. JetBrains, OpenSFF, HighTec, and Turbofish are among the new members of the Rust Foundation. The year 2023 was dedicated to strengthening Rust’s Linux position.
Interestingly, this year, the Ubuntu distribution has been enhanced with the RUST toolchain and all kernel requirements for programming and testing modules (out-of-tree kernel modules) related to the Linux kernel. It is slowly being said that Rust is now mature enough to be treated on par with the C language and thus used to develop selected kernel components. Further confirmation of Rust’s maturity can be seen in its addition to the ‘Safer Languages’ list by NIST in March this year. Despite all this glitz and glory, heaps of work still needs to be done. As we explore Rust-related blogs and forums, we find this particularly true. Undoubtedly, the good news, especially for developers, is the rumor that detailed information on certification and training programs run by the Rust Foundation is expected in the first half of 2024.
The year 2023 also marks the arrival of updates to the MISRA C and C++ standards or the ASPICE4.0 framework. Refreshed after 11 years, the MISRA C version incorporates new rules and directives for the C90, C99, C11, and C18 language versions. We had to wait much longer for the MISRA C++23 standard, as the last version was shown in 2008. The updated release covers C++ language versions up to 17 and is aimed at safety-critical system developers. As with MISRA C, updated rules and directives have been added here too, and industry experts point to some consistency with the C++ Autosar standard. It is not just cosmetically different from ASPICE 3.1, as three significant changes have been made. First, novel process models can be developed, such as those that take machine learning into account in automotive projects or mechatronic device development. The next step involves the evolution of strategy rules. Lastly, there is the revised teaching and training scheme within ASPICE.
I couldn’t help but mention another issue that doesn’t quite fit into the earlier areas. RPI microcomputer enthusiasts can already enjoy the latest RPi5 release. According to the manufacturer, the platform is highly optimized and has better specifications, such as a powerful processor, a newer generation of GPU, and an additional MIPI or PCI connector.
Taking a look at the whole year, 2023 was very dynamic, especially when it came to AI development and cyber security topics. As a counterbalance, the semiconductor industry has calmed down a bit. What will 2024 have in store? We will undoubtedly be astounded by AI development, preferably in a positive way. It is more likely that semiconductors will stay on course for stability. However, we cannot fool ourselves – price increases won’t pass us by, and this will affect the viability of certain products. New languages (or frameworks) are doing well, so I anticipate a steady growth.