Embedded Technology A cross all industries, we are

experiencing a data generation boom, and with increased data comes the need for greater processing

requirements in the growing armada of embedded electronic systems. This is especially pertinent to defence applications, where embedded electronic systems are used for everything from communications to global positioning, on land, in air and at sea. These systems must process input data quickly and effectively, all in harsh operating environments, which involves an ever-increasing amount of processing power and careful design considerations.

As many applications have become more data-intensive, there has been an increase in computing techniques such as graphics processing unit (GPU) accelerated computing to improve system capabilities. This has led to wide use of general-purpose calculation on graphics processing unit (GPGPU) in many industry applications, with parallel computing platforms like NVIDIA’s Compute Unified Device Architecture (CUDA) entering into embedded computing systems beyond the realm of gaming where it was originally used.

GPGPU is most readily used in applications that typically operate in a stable, and often temperature-controlled, environment, like telecommunications. The transition has not been as smooth for computing platforms that are to be subjected to extreme shock and vibration, or extreme temperature and humidity fluctuations that range from sub-zero to triple digits, in addition to supporting crucial, lifesaving and security- focussed applications.

Fortunately, Recab UK’s experience is that these two ideas—of the high processing potential of GPGPU solutions with ruggedisation for harsh operations—can easily be reconciled. We can reliably say that the principles of ruggedisation for embedded systems can be applied to GPGPUs. Our industry partner, Aitech, recently published an application note entitled, “Process High Volumes of Data in Rugged Embedded Systems, No Matter the Requirements,” which outlines some of the steps that can be taken to ensure GPU accelerated computing systems are deployed in the most effective way for defence applications.

System requirements

The best first step is for system designers to take stock of the system. Today’s military systems use far more resources in a much smaller footprint, typically referred to as optimised SWaP, or size, weight and power. This needs to be achieved while keeping costs low. In addition, these applications function in extremely harsh environments, and carry with them the need to operate reliably all the time, every time.

Designing embedded systems for mission-critical data processing

Information and intelligence have long been the backbone of successful military and defence operations. Increasingly, that intelligence is driven by data collected by devices and systems in the air, on land and aboard naval vessels, which requires reliable, rugged embedded systems capable of handling vast quantities of data. Here, Mark Jeffrey, technical director of rugged defence computing specialist Recab UK, outlines the key considerations when designing embedded systems for processing data in defence.

As Aitech says, “this dichotomy has plagued many electronic engineers developing critical military, and defence systems for decades, but these challenges can not only be mitigated, but met as well.”

Crucially, it’s important to always balance system needs and processing requirements with the rugged aspects of the application environment. By relying on proven ruggedisation techniques as well as verified testing methodologies, GPU accelerated computing can offer unique advantages in system performance, even in the harshest of environments. There are three key areas where GPGPU computing provide significant value: processing, memory and power consumption.

Data processing

In mission-critical applications, the accuracy, reliability and speed of embedded systems are essential to a real-time, decision-making process, whether it’s a response required by an actual human or through artificial intelligence. This is especially true of defence applications. Because it is based on a parallel architecture, GPGPU computing processes tens of thousands of data points simultaneously, versus only hundreds using serial processing. Even a typical multicore CPU-based architecture only offers a handful of cores running in parallel. When integrated into a ruggedised system, GPGPUs can meet the growing data requirements of today’s military, defence and space applications.

Memory and storage

One of the challenges that many defence applications face that other industries often

do not is the volatility of connections. Rugged and mobile systems may frequently be unable to access a network to communicate data in real-time. In such an event, we must consider where this data is stored until it can be transmitted. Once it can be, the data must then be communicated quickly and correctly. Onboard storage capacities have been increasing, due to the variety of compact, high density Flash-based modules, high-speed non-volatile memory express (NVME) protocols and secure peripheral component interconnect express (PCIe) based interfaces now available. In addition, memory capabilities have been enhanced by GPGPUs, allowing processing closer to the edge where it is needed most. So, when the data is transmitted, it can already be pre-processed and actionable.

Power consumption

There are two aspects to power consumption that we consider when developing an embedded GPGPU platform for defence. First, managing power resources across compact, high density systems, and second, increasing the power-to-performance ratio of these systems. Some GPGPU boards are very power efficient—especially those based on NVIDIA’s Jetson family, due to the CPU being ARM- based, such as those from Diamond Systems and Aitech—with some boards offering the same consumption requirements as CPU and GPU boards together.

However, GPGPU boards can process far more data using thousands of parallel CUDA cores, meaning a good power-to-performance ratio. More processing is available to the

application for the same, and sometimes slightly less, power.

GPGPU computing is quickly becoming a platform for advanced computing intelligence, due to the hundreds of high-performance cores that provide unprecedented parallel processing capabilities, using general purpose programming languages, such as NVIDIA’s CUDA API. With the proper ruggedisation techniques in place, military, and defence programmes can benefit from GPGPU computing systems. To cite one example from Aitech’s application note, a GPGPU system is invaluable in mobile land vehicles. Today, tanks and other ground military vehicles are often relied upon to send mission critical data from the battlefield to command centres, or directly to soldiers on the ground. As such, it’s quite common for these vehicles to be fitted with multiple onboard cameras and data collection points. Here, a rugged GPGPU system might be responsible for capturing images from six cameras— four composite and two HD-SDI — and then performing simultaneous image processing applied to object recognition and classification, as well as situation awareness. The system is using CUDA for image processing and saving sensitive data on internal fast NVME SSD that can be transmitted back to the command centre instantly and when needed.

As the volume of data produced on the frontline continues to increase, more powerful embedded systems will become a requirement. It’s only by staying one step ahead that military and defence applications can maintain the advantage of intelligence.

Components in Electronics July/August 2020 31

Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20  |  Page 21  |  Page 22  |  Page 23  |  Page 24  |  Page 25  |  Page 26  |  Page 27  |  Page 28  |  Page 29  |  Page 30  |  Page 31  |  Page 32  |  Page 33  |  Page 34  |  Page 35  |  Page 36  |  Page 37  |  Page 38  |  Page 39  |  Page 40  |  Page 41  |  Page 42  |  Page 43  |  Page 44  |  Page 45  |  Page 46  |  Page 47  |  Page 48  |  Page 49  |  Page 50  |  Page 51  |  Page 52  |  Page 53  |  Page 54  |  Page 55  |  Page 56  |  Page 57  |  Page 58