Artificial Intelligence at the edge

MicroSys Electronics is an embedded and edge computing expert that specializes in solution platforms based on NXP® processor technologies. We provide customers with application ready solution platforms on board and system level. To enable the implementation of Artificial Intelligence at the edge, we collaborate with AI processor vendors such as Hailo. On this landing page we want to introduce you to the benefits of combining the technologies from NXP® and Hailo – which as you will see is a match that’s truly made in heaven.

AI for industrial vehicle or robot platforms

Within the broad spectrum of markets for AI ranging from automotive and Industry 4.0 to smart cities, smart homes and smart retail, the bundle of automotive grade NXP® processors and Hailo AI technology is the perfect solution for developers, if they

  • are looking for a headless industrial vehicle or robot platform, which means without HMI
  • need fast data throughput at I/O level
  • need real-time determinism
  • and possibly even address functional safety markets

Typical vertical markets for the NXP® S32G and Hailo AI bundle

 

Typical applications for this NXP® S32G and Hailo AI bundle include, for example:

Agricultural machinery and seed robots
All types of industrial robots

There are countless other applications that can benefit from the S32G processor integrated features, including even stationary devices. And while NXP® is serving the larger automotive OEM and Tier 1 suppliers, MicroSys addresses all applications for this processor that lie beyond standard high-volume automotive applications.

Why is NXP S32G a perfect fit for industrial vehicle or robot platforms?

  • The S32G family from NXP® primarily addresses all those applications that need processor integrated I/O support for vehicle networks
  • S32G processors come with either 4 Arm® Cortex®-A53 cores and 3 Arm® Cortex®-M7 cores or with 8 Arm® Cortex®-A53 cores and 4 Arm® Cortex®-M7 cores as dual lockstep pairs
  • The integrated 18x CAN FD, 2x FlexRay and 4x LIN interfaces can adapt to nearly all vehicle networks that exist worldwide
  • And with TSN support over 10BASE-T1L – which is a simple twisted pair Ethernet connectivity that enables the transport of messages with synchronized time stamps for real-time operations – we are ready for the future of IP based, cloud native communication in vehicles as well

If all this is already embedded in a processor

  • the I/O latency is reduced
  • a lot of the groundwork is already done so you can finalize your building block evaluation on the chip level much faster
  • all this comes with ASIL-D functional safety that can be adapted to various non-automotive markets as well by adjusting the standards

Why do we need AI at the mobile vehicle edge?

The combination of automotive grade processors with AI is needed, for example, to analyze data locally to generate more insights from sensors. The goal is to reduce the heavy communication workloads as connected vehicles can generate terabytes of data per hour. And for real-time functional safety it is of course self-explaining that you don’t want any latency.

AI for situational awareness

All LIDAR and video camera based applications need AI for situational awareness to improve the capabilities for ADAS and autonomous driving. This is one of the largest markets to be addressed beside pure I/O data analytics. Smart digital rear view mirrors are another great application example as well. So there is a huge variety of applications where this bundle of NXP® and Hailo makes sense.

AI enables new business models

With their end applications in mind, customers are also looking for great new appliances coming out of the subscription economy, for instance to offer subscription based on demand maintenance services, or pay per use models to switch CAPEX to OPEX. Even insurances are looking for such data to analyze the risk of operations, and of course, application data are also needed for all closed loop engineering issues.

What are the typical AI applications in vehicle networks?

AI produces situational awareness for various vehicle networking applications such as ADAS, autonomous driving, smart rear view mirrors, big data analytics for predictive maintenance and closed loop engineering, and new business models.

ADAS

Autonomous driving

Smart rear view mirrors

Augmented reality

Closed loop engineering

New business models

What exactly makes Hailo AI so attractive?

There are four good reasons:

  • One of the fastest AI available in the market
  • Extremely low power consumption
  • Available in an automotive qualified shape
  • Scalable and supported by an open ecosystem that can be addressed with the leading AI engineering environments

One Hailo-8 AI processor can, for example, deliver a performance of up to 26 TOPS at a highly efficient power consumption of typically just 2.5 Watt. When you combine this with automotive solutions such as S32G or even Layerscape® processors, you have everything you need for your mobile vehicle or robot platform:

  • Fastest data throughput at I/O level
  • Real-time determinism
  • Processor integrated functional safety tailored to your vertical market needs

What is a typical setup for the combination of Hailo and S32G processor technologies?

With the S32G, you can use the Hailo for example as a smart pre-processor for cameras that are connected via a MIPI camera interface. As the Hailo-8™ has its own on-chip memory, the computation is extremely efficient and fast.

The Hailo performance benchmark

To see how efficient the Hailo-8 is, Hailo has compared some Nvidia platforms in different benchmarks such as ResNet_v1_50 for segmentation or SSD MobileNet for object detection. As the graphic shows, Hailo not only performs better than competitive platforms but the power efficiency measured in frames per second per watt is also significantly better. This makes Hailo a perfect fit for small sized devices such as S32G based systems with a low power envelope. That is very beneficial for many mobile applications!

Acceleration through application ready solution platforms

When it comes to the question of how to get these fabulous chips into a platform that is tailored to specific needs, the most common approach for engineers is to have a look at a higher level of integration based on a modular concept. Hailo fully understands this and therefore offers its AI cores with M.2 extension cards; and MicroSys can deliver the board level products that are tailored to those specific needs. MicroSys is – until today – the only vendor worldwide to provide the S32G family on System-on-Modules.

WeMOTA: Smart analytics of vehicle data all over the world

We have tailored a SBC for instant evaluation that is proven in real world applications such as the WeMOTA platform. Mobile vehicle engineers put such systems in vehicle fleets to analyze working conditions, for example to identify turbo loader failures, which can for instance happen on high hills where the air has less oxygen. Such a comprehensive solution offering from modules to even the most rugged customized systems can help you on any value level – whether you are looking to just purchase the modules, or want to hand over the entire system platform requirements to a strong partner that is part of the very specific ecosystem you need.

NXP® LX2160A based edge server platforms for accelerated packet processing

And just in case you fear that the S32G may be a bit too narrow-chested for your application: Then please have a look at our LX2160A offering. It is a server grade platform with 16 Arm® Cortex®-A72 cores that you can test even in combination with the S32G in the Blue Box. Okay ­– this particular box made available by NXP® as a test platform for automotive edge applications only features our S32G module. But MicroSys can always create platforms tailored to your specific needs to accelerate the design-in for rapid system development with scalable AI in communication heavy edge environments.

Rapid system development with application ready platforms

MicroSys provides two engineering platforms that can be equipped with between one and five Hailo AI modules on S32G and LX2160A based modular processor boards to allow you to instantly test the setup and run your application. They are already booting with Hailo and can be ordered as application ready development platforms. If you want to find out more information about these solution platforms please contact our Sales!

miriac® AIP-LX2160A

High-performance embedded AI platforms

First application-ready building block based on the NXP® QorIQ® Layerscape® LX2160A processor

miriac® AIP-S32G274A

High-performance embedded AI platforms

Second application-ready building block based on the NXP® S32G274A vehicle network processor.

miriac® AIP-S32G399A

High-performance embedded AI platforms

Third application-ready building block based on the NXP® S32G399A vehicle network processor.

NXP meets Hailo-8 webinar

At the Italian Progettista+ virtual conference 2023, MicroSys was invited to present the benefits of the marriage it created between Hailo-8 artificial intelligence and NXP® S32G vehicle network processors. If you’d like to learn more about the benefits for AI supported ADAS and autonomous driving please watch this video.

Contact

Do you have any further questions about our NXP® and Hailo based products or service offerings? We are looking forward to your inquiry and will give it personal attention. Here's an overview of how to reach us.

News

All the latest news about MicroSys, new products and services as well as press releases can be found in our news overview.