Unmanned sensor and processing payloads proliferate

Unmanned aerial vehicles (UAVs), also known as remotely piloted vehicles (RPAs), outfitted with advanced sensor payloads are actively gathering, and even processing, a wealth of intelligence, surveillance, and reconnaissance (ISR) data."There's a proliferation of sensors," says Christopher C. Ames, director of international strategic development at General Atomics Aeronautical Systems in San Diego. The number of unmanned ground, aerial, and undersea vehicles deployed by aerospace and defense organizations has grown exponentially given the many benefits they deliver.

Predator unmanned aerial vehicles from General Atomics Aeronautical Systems are used extensively for combat missions, including ISR activities.

Unmanned benefits

"Sensors are the eyes and senses for the remote [operator], and these sensors feed the processing subsystems," which lend to effective decision-making and active flight management," says Dagan White, aerospace and defense product marketing manager at field-programmable gate array (FPGA) specialist Xilinx Inc. in San Jose, Calif.

A key advantage of unmanned platforms is persistence, says Paul Monticciolo, chief technology officer of Mercury Computer Systems and president of Mercury Federal Systems Business Unit in Chelmsford, Mass. Manned missions can be limited by human endurance factors, whereas unmanned aerial vehicles (UAVs) can surveil the area of interest for 24 hours or more.

"Pilot safety is an important aspect," Monticciolo adds. "It helps remove the human element. For ISR, unmanned platforms have shown exceptional utility."

"The benefit of combining an unmanned aircraft and sensors is the flexibility in deploying and operational use of those sensors in a desired scenario," Ames explains.

General Atomics equips its unmanned aircraft with a variety of sensors that make it a very effective tool in generating persistent situational awareness. The company's Lynx synthetic aperture radar provides all-weather, wide-area surveillance, takes high-resolution, photographic-quality radar images, boasts a ground-moving target indicator (GMTI) capability, and can see through clouds, Ames says. "There's no point in having an ISR platform if you're unable to penetrate clouds with a sensor. It also harkens to the need for flexibility in sensors, so it has a maritime capability. When you combine that with an electro-optic infrared (EO/IR) capability for days when weather is clear, you have a formidable ISR collection suite.

"Using [CLAW] software integrates all sensors on an aircraft. It makes the sensors operate complementary to one another. For example, you can pick up a Lynx synthetic aperture radar image and, with a series of points and clicks, the software will automatically redirect the EO/IR device to point toward and lock upon the radar image so that you can cross-queue and determine if it is a contact of interest," Ames adds.

Added advancements

Technology firms continue to answer the call for greater unmanned payload capacity and more advanced sensor payloads. "Ground-based robots have proved their value in the field many times over, being able to venture into places where a human would be at great risk," observes Peter Thompson, system architect, GE Intelligent Platforms in Towcester, England. "Dealing with improvised explosive devices (IEDs) and leaking nuclear reactors are but two examples.

"There is a drive to improve their sensor suites: higher-resolution cameras, fusing EO and IR sources, electronically removing jitter from moving images, and so on," Thompson adds. "Wide-area surveillance platforms are being fitted with more sensors of higher capacity. An ISR platform might combine electro-optical, infrared, and synthetic aperture radar modalities to provide better coverage in differing conditions.

"Each sensor type is being continually improved with bigger focal plane arrays, faster frame rates, and bigger swaths," Thompson says. "One piece of the system that struggles to keep pace is the bandwidth of the communications link. The usual analogy is that of a fire hose of incoming data being connected to a soda straw for output."

"There is a seemingly insatiable demand for bandwidth and processing capability in UAV platforms, but it must be balanced against the need for reduced power consumption to enable extended mission capabilities," White says. "Full-motion video feeds are crucial for remotely based pilots, and this further drives the overall bandwidth demand."

The only practical way to approach the problem is to move more and more of the signal, image, and data processing onto the platform, Thompson advises. "This has to be done within strict constraints of size, weight, and power (SWaP) to minimize effects on the overall performance of the platform in its mission."

GE is harnessing the power of supercomputer installations and replicating it in a package that is small enough to fit where it has to go and robust enough to survive extremes of temperature, shock, and vibration, Thompson says. "Supercomputing and high-performance computing (HPC) is now available in the form of high-performance embedded computing (HPEC). HPEC is fundamental to managing sensor payloads on unmanned vehicles because of the high bandwidth of data, processing complexity, need for very rapid response times, and pressure on SWaP."

GE, working with Juniper Networks, has introduced the RTR8GE secure battlefield router to improve mission-critical communications and information sharing.

Bad bandwidth

Mobile platforms, such as unmanned vehicles, often are connected to the global network through a relatively low-bandwidth link, explains Eran Strod, system architect at Curtiss-Wright Controls Defense Solutions in Ashburn, Va. "The traditional workflow required the data to be transmitted back to a central location where commercial high-performance computing (HPC) systems would perform back-end processing. Then, an analyst would physically look at the information of interest in order to determine any desired actions. The delay in this workflow is several hours or even days. Unfortunately, the information was often obsolete by the time it reached those who needed it."

Moving processing to the mobile platform, near the sensor, could enable faster response times but requires an HPEC system, Strod says. "It's a similar processing model, but performed on the data in real time without the bottleneck of the low-bandwidth link back to a data center. Automated algorithms perform first-order processing and transmit only select imagery and other metadata over the link. This workflow is much more optimized and gets much closer to real-time response. The key is putting rugged processing elements adjacent to the sensor where there are no bottlenecks.

"The best way to achieve supercomputing performance at the sensor today, by far, is with a combination of Intel processors and general-purpose graphics processing units (GPGPUs)," Strod adds. "These processing elements tend to consume a fair amount of power so air-flow-through technology (VITA 48.5) is emerging as a significant enabler."

General Atomics CLAW software can be con-figured to control EO/IR, synthetic aperture radar, and data link payloads simultaneously.

Onboard processing

Aerospace and defense end users typically must adhere to certain standards for data-link communications. Whereas the commercial industry can evolve communications quickly, such as rapidly moving from 3G to 4G, the size of the military installed base is such that upgrading is too expensive, Monticciolo admits. "One way to deal with that problem, which is not going to be solved for quite some time, is to do much more processing and exploitation on board.

"What I expect to see in the future is more of the autonomous type of processing where you're doing not just detection, but also tracking, targeting, determining target characteristics from the data, and passing along only key information directly to the warfighter," Monticciolo predicts. "That's the vision. We have this constriction of bandwidth and we have to figure out smart ways to deal with it and get that actionable information and intelligence directly into the warfighters' hands. Putting more and more of the processing on the UAVs and sending down the essential data over those limited links is really the path to the future."

Mercury Computer's processing and storage technologies are playing a role in current and next-generation unmanned systems. Increment 1 of the Gorgon Stare program, a wide-area electro-optical and infrared sensor system on a General Atomics MQ-9 Reaper aircraft for the U.S. Air Force, involves a Mercury processor and storage system tied to ITT Exelis EO and IR cameras and an L-3 Communications system. Sierra Nevada Corp. is the prime contractor on the Gorgon Stare program.

The Gorgon Stare payload looks at all the information available in "a moderate-sized city diameter," Monticciolo explains. "We are taking information from either of those camera types and [putting it] through the image-rendering process. In addition to providing that full operating environment, you have the ability to send directly to the warfighter small segments of information in real time."

Gorgon Stare has been deployed in theater. The next phase involves digital cameras with enormous data rates that need to be processed, for which Mercury engineers are leveraging commercial technology with a combination of GPGPUs. "The gaming industry is where these GPUs have come to fore," Monticciolo says. "We've taken those processor capabilities and applied them to sensor signal processing. We can do advanced image processing in a very power-efficient manner using these particular chips."

Mercury Computer also employs field-programmable gate arrays (FPGAs) to handle the high data rates coming in from sensors, as well as to perform image compression and other signal processing functions. Company engineers also "take server-class microprocessors from Intel and package them appropriately [with] efficient ways of removing the heat so that we can put the same processors that are in the servers you find in ground stations and put them right on the [unmanned] platform," Monticciolo enthuses. "Now, you could wind up taking the same software and same kind of exploitation processing you do in ground stations and do that onboard the platform."

Mercury Computer also provides "solid-state data storage subsystems, which are very important in exploitation because you want to exploit databases for maps, past behaviors, etc.," Monticciolo says. "How do I correlate the information I have obtained now with past behavior I have observed? The ability of Mercury processing on this Gorgon Stare platform to ingest data, process it, perform exploitation, store it, and disseminate it to the ground and ground stations is extremely powerful."

Compute power

Users of unmanned vehicles and sensor payloads will "continue to push the limits of compute power bundled with more sensor data acquisition capability," but not to the detriment of the payload size, predicts R.J. McLaren, marketing manager, military products at Kontron in Poway, Calif. "SWaP comes at a premium in these types of vehicles. We certainly expect customers to want to increase the processing capability in order to analyze data locally, which will allow them to make real-time decisions. This may actually cause the need for more payload capability, as those decisions may result in additional requirements for other capabilities."

Kontron delivers board- and system-level products-including 3rd Generation Intel Core i7 processors, high-performance GPGPUs, switches, and carrier boards-to the unmanned vehicle market, including standards-based, conduction-cooled solutions that engineers can leverage to build rugged systems. Engineers working on unmanned vehicles also tap the company's configured systems, with boards enclosed and loaded with the operating system, drivers, and BIOS configurations.

"On the storage side, solid-state drives (SSDs) in the required extended environmental range continue to grow in capacity and features, like encryption and secure-erase capabilities," McLaren describes. "Our Cobalt and ApexVX conduction-cooled systems are optimized for SWaP and are a standards-based modular design that can support a wide range of requirements. This market continues to see an increase in requirements as engineers innovate how unmanned vehicles can be used and as there continues to be a successful track record with deploying these systems."

Parallel processing

Unmanned vehicle sensor payloads continue to capture a seemingly insurmountable volume of data, including high-resolution imagery and full-motion video. Concurrently, solutions providers are harnessing the latest innovations, including commercial off-the-shelf technologies.

"Incorporation of preprocessing in the FPGA fabric taking advantage of massively parallel computing power to reduce the data and find the key information of interest, and then post-processing in the ARM processing subsystem for image fusion, manipulation, and export for display is absolutely a strong fit for Xilinx within UAV payloads," White says. Preprocessing and data reduction are key to reducing bandwidth needs for handling outbound data, and are a fit for the Xilinx EPP product line, including Zynq, he says.

Massively parallel computing-taking computing tasks and breaking them down into constituent elements that can be processed concurrently-is the key to success, Thompson recognizes. "Perhaps the most powerful technology at our disposal is the GPU, because its architecture specifically allows for a very high degree of parallel processing. GE is deploying arrays of GPUs into the processing of large imaging sensors to handle the huge amount of data coming in and to detect and track items of interest in systems that are much smaller than would be the case with general-purpose processors. NVIDIA's latest-generation Kepler GPUs, for example, feature 384 cores; that's like having 384 processors working in tandem, yet they occupy the space of just a single processor."

Data compression technology also holds promise in reducing the size of sensor-acquired data files. GE engineers are using compact image compression devices to reduce sensor data to the point where it can fit a restricted bandwidth data link on small platforms, Thompson reveals. "We have replicated the functionality of large racks of commercial servers on the ground in compact systems that can fly, yet are software compatible with the ground-based systems and are powerful enough to fuse the data from different modalities."

Safety and security

Downed UAVs, although unfortunate events, demonstrate the importance of safety features in unmanned systems, White recognizes. "Safety and reliability are imperative to the future success and exploitation of UAV technology. Secure communications links are vital for the UAV," and a continued area of focus for Xilinx.

"Single event upset (SEU) must definitely be considered in any avionics system," White recommends. "At UAV altitudes, the upset rate is hundreds of times greater than at sea-level. It is critical to consider this when designing for safety and performance." Xilinx offers radiation testing, publicly available data, and supported SEU mitigation IP and tools to aid systems designers. "It is absolutely imperative to address the SEU issue early in the design cycle to ensure that the system is architected for maximum performance and availability. If you wait until the end, you may end up with a sub-par system design."

"The move to network-centric systems-for example, the VICTORY architecture-and use of cloud processing is leading to a need for much better network security in the field," Thompson explains. "Demand for firewalls and routers with intrusion detection and prevention is increasing." To meet such needs, GE entered into a relationship with Juniper Networks, and introduced the RTR8GE secure battlefield router.

Future functionality

Persistent situational awareness provides a knowledge advantage and validates the value of unmanned aircraft, says Ames. "Sensors with wider areas, faster processing, and on-board processing to save bandwidth-that's the future."

"The proliferation of Intel CPUs in DSP systems is bringing software programming environments, tools, and middleware used in the commercial markets to embedded sensor processing," Strod says. "Applications written on a software stack will much more easily port to future processors in a tech refresh, and there will be much more sharing between programs, lowering cost and increasing the pace of innovation."

Unmanned sensors will be an area of both evolution and revolution over the next 10 to 15 years, Monticciolo says. "You're going to be able to use the advances of electronics to extract more information out of the environment, process it, and bring it onto the ground."

COMPANY INFO

AdaCore
www.adacore.com

AeroVironment
www.aerovironment.com

Aeryon Labs Inc.
www.aeryon.com

Cobham
www.cobham.com

Curtiss-Wright
www.curtiss-wright.com

FLIR
www.flir.com

GE Intelligent Platforms
www.ge-ip.com

General Atomics Aeronautical Systems
www.ga-asi.com

General Micro Systems
www.gms4sbc.com

Goodrich
www.goodrich.com

IAR Systems
www.iar.com