Intelligence — about enemies, terrain and everything in between — is an essential aspect of military operations, and it’s driving innovation in data collection and processing at the tactical edge. However, developing rugged embedded systems that meet the operational and environmental requirements of modern combat is far from a simple task. Here, Martin Frederiksen, managing director of rugged defence computing specialist Recab UK, explains what goes into designing rugged systems that are fit for modern battlefields.
Renowned military theorist Carl von Clausewitz famously noted that two basic principles underlie all strategic planning: “to act with the utmost concentration” — that is, to identify enemy centre of gravity and coordinate actions on them — and “to act with the utmost speed”. These principles remain as valid for today’s military operations as they were when first written in the 19th century — so much so that they are reflected in the deployment of many modern military technologies.
Embedded computing systems feature on the frontline of many military and defence applications, from video surveillance processing on military drones or relaying position coordinates on military vehicles. Many of these technologies serve the purpose of collecting and sharing information, be that collecting data from the area of deployment, supporting communication between soldiers and bases, or sharing positioning data to coordinate activities. Each of these help forces to identify enemy activities or bases and organise their actions accordingly, with increasing speed.
The potential for newer technologies to better serve the military’s ability to gather and collate intelligence efficiently is central to many military modernisation initiatives globally. In the UK, the recent Defence and Security Industrial Strategy set its sights on “[building and sustaining] a ‘digital backbone’ to share and exploit vast amounts of data, through the cloud and secure networks”, according to UK Defence Secretary Ben Wallace.
A similar initiative is in place in the US in the form of the DoD’s Joint All-Domain Command and Control (JADC2). In a Congressional Research Service report in March 2021, the JADC2 was described as a concept that would connect sensors from all of the military services into a single network. The report states that the DoD “envisions creating an ‘internet of things’ network that would connect numerous sensors with weapon systems, using artificial intelligence algorithms to help improve decision making.”
Overhauling and modernising operations as complex as those in military and defence is no easy feat due to the requirements of the embedded systems at the tactical edge. These systems must first process input data efficiently and effectively in harsh operating environments, which involves an ever-increasing amount of processing power and careful design considerations. The processed data then needs to be communicated rapidly, reliably and securely to other networked systems.
In military operations it is not only the quantity of data that has increased in recent years but the quality, which in turn increases the required processing capabilities. A prime example of this is in military unmanned aerial vehicles (UAVs) used to survey battlefields or execute attacks. Drones used in the first decade of the Global War on Terror were notoriously limited in the quality and size of video feeds, with former US Air Force pilot Timothy Cullen likening them to “looking through a soda straw”. Today, the image quality and visible area provided by drones is markedly improved — but more process intensive.
To handle the increase in data intensity, embedded computing has seen a rise in the use of Artificial Intelligence (AI) and General-Purpose calculation on Graphics Processing Unit (GPGPU) to accelerate computing capabilities. By relying on proven ruggedization techniques as well as verified testing methodologies, GPU accelerated computing can offer unique advantages in system performance even in the harshest of environments, notably in terms of data processing, memory and storage, and power consumption.
AI and GPGPU computing is based on parallel computing architectures, such as NVIDIA’s Compute Unified Device Architecture (CUDA). This allows GPGPUs to process tens of thousands of data points simultaneously, versus 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 applications in terms of raw processing capacity at the edge.
With the data pre-processed or locally processed, it then generally needs to be communicated. However, unlike more stationary operating environments, rugged, mobile and mission-critical systems operate in places with volatile connections and may need to house data onboard until the network is restored. As such, storage is a key consideration. Fortunately, developments in compact, high density flash-based modules, high-speed NVME protocols and secure PCIe-based interfaces have increased onboard storage capacities alongside the edge processing potentials of GPGPUs reducing the data footprint.
These factors explain why AI GPGPU-based computing is presenting numerous opportunities for military and defence OEMs to develop highly sophisticated edge systems. For rugged SWaP-optimised systems that will need to capture graphics from several HD-SDI/composite video inputs and manage data from many I/O interfaces, while providing great image processing capabilities, a rugged AI GPGPU-based system using CUDA parallel computing is the ideal solution.
It’s for this reason that, in many of the recent military projects that Recab UK has undertaken, we leverage our close partnership with Aitech to provide robust, rugged AI GPGPU based solutions for extreme environments.
Yet the potential of GPGPU and AI at the military tactical edge not only supports the modernisation efforts of forces globally. It is also accelerating the capabilities of entirely new systems.
For example, prototypes of autonomous electric vertical takeoff and landing (eVTOL) aircraft are in development, using platforms that already exist, such as drones or unmanned helicopters, and then integrating leading edge technologies to achieve the needed functionality. Rugged GPU accelerated computing is at this forefront of this. In fact, the technology is advancing so quickly that systems are moving to next gen architectures as development is taking place. In this instance, accommodating the increased sensor processing integrated into the unit is cause for the upgrade to replace typical CPU-based embedded computing architectures.
Even with the most effective and efficient collection and processing of data at the tactical edge, military embedded computing still faces its biggest challenge in the form of sharing that data reliably and securely across a distributed network. Such networks require provide low-latency, high-availability, interoperability between networked devices and high-bandwidth to be effective.
However, many of these networks make use of Ethernet for networking due to its generally lower pricing, ubiquity and standard performance. But because Ethernet is traditionally non-deterministic in nature and designed to function in a way that relies on network loads, it is not suitable for non-static field operations that need high degrees of synchronisation or benefit from real-time data transmission.
New standards are emerging to meet the mission-critical demands of military networks. Among the most notable from a military perspective is time-sensitive networking (TSN), a set of standards under development by the Time-Sensitive Networking task group of the IEEE 802.1 working group. TSN sits at the data link layer of the OSI model and provides determinism to networks. This determinism enforces advanced quality-of-service policies, while also supporting the merging of real-time, reserved and best-effort traffic in the same network. It is also unique in that its streams are delivered with guaranteed bandwidth and deterministic latency.
Several IEEE standards under development that outline specific features of TSN. Although not all standards are required to support TSN, networks and switches that can meet select standards offer a clear advantage in military environments. For example, IEEE 802.1CB outlines the addition of frame replication and elimination, which improves the availability in TSN networks and offers seamless redundancy.
Arguably the most important for military networks is the IEEE 802.1AS synchronisation standard. TSN’s synchronisation is based on the IEEE 1588-2008 standard precise time protocol (PTP), which allows networked devices to share the same time reference within a nanosecond time range. This means it is possible for Ethernet networks to provide a level of synchronisation comparable to GPS.
Using the TSN protocol involves the use of compatible Ethernet switches in embedded systems. Of course, the switch itself must also be ruggedised to handle the operating conditions and certified to military standards including MIL-STD-810G & MIL-STD-461G, while also offering sufficient data speeds. One manufacturer leading the market in this regard is SoC-e, who provides the Relyum range of military commercial off-the-shelf (COTS) managed 1/10G ethernet switch, router and edge computing equipment.
The RELY-MIL-SWITCH-ROUTER platform supports up to 20x 1G copper and up to 6x 1/10G fibreoptic ports, with support for different media types and its distribution in the MIL-DTL-38999 connectors allowing for complete and cost-effective network infrastructures. Relyum equipment also supports several security functionalities, which help to achieve the strictest requirements in military applications.
The concept of fully connecting frontline technology supports the principles of effective strategic military planning for modern technological battlefields, so it is no surprise that forces globally are moving towards realising it. The key to success is ensuring that embedded systems at the tactical edge are up to the task — and this can only be achieved by giving due consideration to the environmental, processing and networking conditions. Fortunately, the technologies are emerging, ready for deployment.