embedded award 2023: And the winners are …
A top-class jury has evaluated 98 submissions – a new record – for the 19th embedded award. "I was impressed this year by the continuingly high level of innovation in the industry and by the extremely wide scope of the submissions," emphasizes jury chairman Prof. Dr.-Ing. Axel Sikora. We present the winners in the eight categories …
Outstanding innovations in the field of embedded system technologies
Artificial intelligence (AI) is one of the two new categories of the embedded award 2023 to follow the exciting trend of more and more intelligent systems in the embedded edge. It is interesting that the predecessor event to embedded world back in the 1990s was called “embedded intelligence”. What was rather a dream at that time, has become pretty real in these days. And of course, we should be reminded that the embedded AI technologies are still in infancy – despite all the impressive progress.
Interestingly, it is not the old paradigm any more to learn in the cloud and to infer in the edge, but the topic is increasingly about learning in the edge, thus reducing the amount of network traffic, increasing privacy, and enabling real-time characteristics.
In this vibrant field, it was very difficult to select the most promising solutions.
The first price in our category “embedded AI” goes to a nice use case and demonstrator and thus to AITAD, who are exhibiting at the booth of Silicon Labs.
They presented a “AI IR shower monitoring sensor wireless module”, which allows the monitoring of activities under a shower – in order to increase the safety for example for elderly people, to save water and energy or to adapt the shower characteristics to the given situation.
In any case, the more precise, low-fluctuation and situational control is possible, since body points can be focused on and evaluated at the detected persons.
Until now, all this was neither cheap nor wireless and privacy-compliant (i.e., not camera-based) feasible. The AITAD demonstrator is based on a low-resolution IR grid sensor, thus addressing privacy concerns in the shower, starting at 8x8 or 16x16 pixels up to three-digit pixel numbers.
Furthermore, it is enabled by AITADs embedded AI process with proprietary conversion algorithms to the AI/ML acceleration unit of Silicon Labs' EFR32-SoC/MCU, which is based on a Cortex-M33 core. The module can be connected directly via LIN and comparable data buses or wirelessly via the integrated 2.4GHz RF unit.
Increasing softwareisation does not change the fact that hardware platforms are the basis for every embedded system. This category having once more the most submissions for the embedded award this year underlines the importance of embedded hardware products. Especially three nominees convinced our jury. And it will be obvious that hardware is far from being “just commodity”.
And the winner in the hardware category nicely is a startup. We will touch the significance of startup companies for the embedded world again later, when we are talking about the startup category.
The company GreenWaves Technologies is awarded the first prize in this hardware category for its GAP9 Processor.
GreenWaves’ GAP9 is a completely new type of processor designed from the ground up to address the requirements of next generation hearable products that will offer a new level of audio quality, adapting themselves to their wearer and their wearer’s environment. To meet these requirements, GAP9 is an ultra-low power and easy-to-program processor that is able to process classic and ultra-low latency Digital Signal Processing (DSP) and Neural Network workloads in highly energy-constrained devices.
Recent MLCommons MLPerf Tiny benchmarks results showed that GAP9 delivers industry-leading performance on a range of neural network tasks achieving the lowest latency and maximum energy performance of all entrants.
As a basis, GAP9’s use of the flexible RISC-V ISA along with its very homogeneous architecture allows to provide a modern development environment that greatly improves programmer productivity.
Embedded vision systems are posing highest demands to many of the system elements. As a matter of fact, data volumes and data rates are significantly higher than for many other embedded systems. Consequently, demands not only on hardware elements like interfaces or memory, but also on software and AI algorithms are driving the entire industry.
The first price in this category goes to Analog Devices for its product ADTF3175, a one Megapixel Time-of-Flight Module.
Today, many depth sensors use triangulation methods (i.e., stereo 2D cameras or structured light) but this has a number of limitations as compared with accessing true 3D depth measurements. However, the density of such flight sensors has been traditionally limited to VGA resolution.
The product from ADI advances the accuracy and quality of depth imaging such that autonomous machines can operate more efficiently and at a faster rate through the use of a one-megapixel sensor, which is – to our best knowledge the first high-resolution, industrial quality, indirect Time-of-Flight (ToF) module for 3D depth sensing and vision systems available in the market. Thus, the product is able to resolve 3D images with an accuracy of +/-3mm over the full depth range of 20cm up to approx 4m.
Intensive networking, increasing autonomy of devices, higher demands with regards to reliable functions lead to growing demands on the functional safety of embedded systems in hardware and software and on protection against external attacks, on the security.
The first price in this category goes to Infineon Technologies for its product OPTIGA™ TPM SLB 9673.
Trusted Platform Modules are around for many years, to give “things” a unique identification so they can connect to the IoT. This identification can be used to track IoT devices and equipment on the networks, and to validate their access rights. To avoid the risk of counterfeit, this number is protected from being modified.
The special feature of Infineon´s latest TPMs is in the support of not only using elliptic curves (here NIST-P521) but also by so-called XMSS signatures, which is a digital signature scheme being quantum resistant and already standardized by NIST and other organizations. Thus, it provides enhanced classical cryptographic capabilities and is a milestone in the push towards PQC with a quantum-resistant firmware update mechanism.
The design of integrated circuits (ICs), System-on-chip (SoC) and of circuits that are then integrated into ICs, the so-called intellectual property (IP), is a core discipline of embedded system development. This core discipline has now finally found its way into the embedded award as a category – in order to strengthen the visibility of this very advanced design discipline.
The first price goes to MIPS for its MIPS eVocore P8700 multiprocessor system.
Together with the rise of RISC-V popularity, there is a wide range of RISC-V CPUs available. Based on these steppingstones, key enablers are silicon-proven processors, security, and optimizations for specific applications such as support for automotive functional safety.
The MIPS eVocore P8700 multiprocessing system brings a new level of performance to RISC-V. It is a high performance, scalable RISC-V IP core – the first with Out-of-Order (OoO) processing and coherent multi-threaded, multi-core and multi-cluster scalability.
With this approach, clusters of multi-threaded, multi-core CPUs – eVocore processors and other accelerators – can be combined in configurations to balance performance and power consumption.
Also, foundational technologies for specific markets are key differentiators. This includes proven robust safety capabilities for ISO 26262 ASIL-B(D) and ASIL-D systems. The P8700 is also the first CPU in silicon with support for automotive-grade Linux.
Of course, the softwareization is driving the embedded world. And more and more software-defined architectures and devices are entering the market. Thus, we received the second most submissions for the embedded award this year in the software category.
The first price goes to Emproof for its product Emproof Nyx.
It’s well known that large computer and networking systems, servers, and data centres use a variety of industry-standard techniques to protect themselves from cyber-attacks. However, embedded systems can be particularly vulnerable to these attacks because they often have limited resources and fixed functionality.
The security software, Emproof Nyx, which is 3 orders of magnitude smaller than existing solutions implements multiple design features that make it the ideal solution for embedded devices based on ARM, Intel and RISC-V ISA’s.
No access to source code is needed as the transformational software acts like a post compilation step and just requires access to binary code or libraries which also means that it works with any build processes and all existing toolchains. The performance impact is minimal as the developer can choose to protect the entire software or just the critical modules which allows the performance to be customized depending on the resources available.
Emproof Nyx has been developed and written such that it will be certified for the Automotive Functional Safety Standard ISO26262 and the Avionics Software Tool Qualification DO330 No other security software solution designed for embedded devices can offer protection against both IP Theft and Exploit Mitigation in the same package enabling companies who have limited or no security expertise to offer levels of security never previously achievable.
It is to be mentioned that Emproof is a startup company. And the fact that it wins the first price in the busy software category shows the high innovation level of startups in the embedded world.
Young companies are driving technological development. The startup scene in the embedded industry is correspondingly lively and since several years, the embedded world runs a dedicated startup zone. In addition, it is extremely impressive to observe the intense collaboration between multi-billion corporations and small and young enterprises. All this is reason enough to dedicate a separate category to start-ups at the embedded award.
The first price in the startup category goes to Epishine for its product OneCell.
With this price, we are coming back to the field of Energy Harvesting to solve the issue of disposable batteries.
To this end, the Epishine OneCell was developed as an organic solar cell tuned to indoor light with world leading light energy harvesting performance, efficiently converting light to electricity even at very low illumination. The OneCell is a sustainable method for powering small wireless electronics. With a flexible design that can be adapted to intricate and customized layouts to meet specific surface finishes of the product, for example, soft touch plastics, leather, brushed etc.
The unique composition of the OneCell enables cut-outs in the surface without limiting the performance, which allows new light powered applications.
Automated tools increase efficiency and reliability at all levels of the embedded value chain. There are tried and tested tools that have been used for years. However, in order to further optimize development processes, the need for new tools is always there.
The first price in the category tools goes to the company Slint for its product Slint.
Also, this product features Rust as a “memory-safe” and efficient language – in this case for a commercial-grade Rust-based user interface toolkit that could be used to create smartphone-like graphical user interfaces on embedded devices. Rust, produces secure code and, crucially, doesn't worsen performance to achieve it. The language has been steadily gaining adherents and is now at a turning point with it being an officially recognized and accepted language in Linux.
Compared to other Rust-based user interface toolkits, Slint’s focus on embedded and desktop platforms makes it a scalable solution – being able to run on devices with a few KBs of RAM to rendering complex User Interface on desktop.
Slint leverages declarative programming style to allow users to define the user interface in a markup language which can be integrated into an existing code base written in any programming language.