As visitors start convening in Nuremberg for embedded world, there’s a perfect storm brewing in embedded systems development, with the parallel rise of cloud native computing coupled with growth in intelligent, connected IoT devices and edge technologies.
It was paraphrased well in a recent briefing with Remi EL Ouazzane, president of the microcontrollers and digital ICs group at STMicroelectronics commented, in which he talked about the rise in ‘cloudified’ devices. He was referring to the fact that more and more connected IoT devices are increasingly moving intelligence from the cloud to the edge, with a continuum of compute capability between the cloud and the edge.
The theme of this year’s embedded world 2022 conference in Nuremberg is a reflection of that trend: “intelligent.connected.embedded”. Conference sessions include a look at this theme of the cloud to edge continuum. For example, in the first edge computing session on day 1 of the conference features the following talks:
- Thomas Rosteck of Infineon Technologies addresses important aspects of innovation at the edge: specifics of trusted IoT systems, the convergence of software and hardware, the evolution of algorithms in AI, and upcoming new post-quantum cryptography algorithms as the answer to the challenges of quantum computers’ world.
- Channa Samynathan of Amazon Web Services (AWS) talks about architecting embedded edge devices and scaling for complexity. He will discuss best practices in design and manufacturing, provisioning, communication, ingestion, analytics, and application layers and how a properly designed edge architecture allows complex scaling scenarios.
- Chee Hoo Kok, a cloud software engineer at AIOBench, an Intel-funded startup venture, presents research on the effect of simultaneous multithreading (SMT) towards the performance of cloud workloads; enabling or disabling SMT when hosting workload components will result in different workload throughputs. He presents categorizations of these based on the utilization patterns of CPU, memory, disk access and network traffic.
- George Grey of Foundries.io presents on migrating to cloud native solutions for embedded software development and deployment, including initiatives for open source solutions, plus open architectures such as SOAFEE and SystemReady from Arm. Topics include cloud native and DevOps solutions for PKI-based device and fleet security infrastructure, OS solutions for different industry segments, device and fleet deployment and orchestration, and ways of reducing costs of ultra-long term maintenance.
- Flavio Bonomi of Lynx Software Technologies delves further into driving the push from the cloud to the edge. His paper discusses how implementing any functionality sensitive to issues in edge solutions represents an incorporation of requirements typical of embedded computing (security, real-time and safe, deterministic behaviors), into modern networked, virtualized, containerized lifecycle management and data- and intelligence-rich computing.
- David Beamonte of Canonical shows how to build a video analytics application on an edge device running Ubuntu core, using Intel’s OpenVino model. The goal is to train the model to detect specific objects, do some local analysis and send that information to the cloud to take further actions. This is an application that is very useful in many applications such as retail (i.e. detecting articles or counting people), logistics (packet identification), or manufacturing (i.e. keeping track of the quality and number of the products manufactured).
This is a tiny snapshot of the many papers among multiple sessions over the three-day conference. The exhibition itself is spread across six halls, and we’ll bring coverage from the show, but here are some highlights pointed out to us by a selection of the exhibitors.
Foundries.io said its’ FoundriesFactory solution is featured in a variety of live demonstrations in a fleet of IoT devices, including the German-designed unu scooter and the Tailos robot cleaner. The unu scooter is a road-going, two-passenger, battery powered e-scooter; FoundriesFactory is used to deploy a secure, customizable Linux-based software platform with fleet management services in the scooters, giving unu the ability to securely update, manage and maintain their scooters with new functionalities via over-the-air (OTA) updates. Tailos is similarly using FoundriesFactory in its robot cleaners, giving the company fine control over the lifecycle of OTA updates and the ability to optimize build to deployment pipelines with integrated CI/CD for both the software platform (Linux microPlatform) and its own applications.
SECO will demonstrate how to add intelligence to edge devices with its Clea software platform which connects edge electronic devices with the cloud and facilitates real time device monitoring, analytics, infrastructure management, predictive maintenance, and secure remote software updates. Clea combines AI, IoT, and edge and cloud computing with customer-centric services and hardware solutions that can be off-the-shelf or tailor-made. The company said that with Clea, any device can be turned into a cloud-managed intelligent device, thus allowing smart control and monitoring, and gaining actionable, real-time insights using machine learning and artificial intelligence.
Blaize is demonstrating its Blaize graph streaming processor (GSP) architecture with partner companies at the show. The Blaize partners showcasing edge AI solutions based on its architecture are: 1) 7StarLake, with multi-camera object detection streaming into an IPC enabled by the Blaize Xplorer X1600P PCIe accelerator, including a demo of video surveillance as a service from Blaize partner Xompass. 2) Innovatrics, with an edge ready-to-deploy facial recognition technology powered by Blaize Pathfinder P1600 embedded system on module (SoM). 3) Aetina, with a static display of the Blaize Xplorer X1600E EDSFF small form factor accelerator platform for accelerating AI applications at the edge.
NXP Semiconductors will debut its new MCX portfolio of microcontrollers for smart homes, smart factories, smart cities and emerging industrial and IoT edge applications. It includes four series of devices built on a common platform and supported by the MCUXpresso suite of development tools and software. The portfolio features the first instantiation of NXP’s new, specialized neural processing unit (NPU) for accelerating inference at the edge, delivering up to 30x faster machine learning throughput compared to a CPU core alone. The portfolio is based on high-performance Arm Cortex-M cores, integrated with a set of peripherals and featuring up to 4 MB of on-chip flash memory, low power cache and advanced memory management controllers, plus up to 1MB of on-chip SRAM to further enhance real-time performance of edge applications.
eInfochips is showcasing its edge AI inferencing solution, developed in collaboration with Qualcomm Technologies. Key components of this edge solution comprise a camera for video capture, edge computing for face detection, and Amazon Kinesis video streams (KVS) for live streaming and alert generation. Amazon KVS makes it easy to securely stream video from connected devices to AWS for analytics, machine learning (ML), playback, and other processing. The AI inferencing solution uses eInfochips’ camera reference design kit (RDK), built on the Qualcomm QCS610 platform, showcasing the premium high-performance processor’s capabilities for delivering on-device edge AI for advanced applications, including smart kiosks, industrial IoT, AI surveillance cameras, and AI edge appliances.
At embedded computer level, Cincoze said it is making its debut at embedded world, with edge computing solutions for intelligent manufacturing in three areas: rugged embedded fanless computers, embedded GPU computers, and modular panel PC and industrial monitors. In ruggged computers, the DV-1000, launched in May, features a high-performance and compact design with wide temperature support (-40–70°C), is equipped with an Intel Core i-series processor, DDR4 2666 MHz memory up to 128 GB, and has the most essential I/O ports for smart manufacturing.