In recent years, the aerospace industry, particularly in the military market, has seen the proliferation of UAV in a variety of end-use applications. The applications now include everything from long range 30-hour flights at high altitude, to ultra short-range tactical vehicles launched from the shoulder. Concurrent with those advances, and in many cases preceding them, have been unmanned ground vehicles and marine vehicles. As might be expected, the demands on the payloads for these unmanned vehicles has grown exponentially in numerous ways including compute power, ruggedness, restrictions on energy dissipation, flexibility and size minimization. All of these demands tend to oppose each other, making the final solution that much more difficult. The result has been the need for novel approaches to solve the conflicting demands of this market.
GE Fanuc provides the electronics inside the electro-optic (EO) system aboard the Sperwer UAV (Figure 1), a widely used European UAV. GE Fanuc’s board handles target detection and video tracking aboard the Sperwer. The Sperwer currently flying uses a relatively older product that’s in a 3U VME format.
Old Way: Piece by Piece
Solving the challenges noted above can typically be achieved effectively on a per-component basis. That is, a device to perform a task can be engineered to meet certain requirements. That’s been the traditional role of the embedded computer vendor, who would typically be unaware of what the final system topography was going to be. To build a UAV payload system, a system integrator would typically contact one or more embedded computer suppliers and obtain specification data sheets.
Elements of a system would include items such as enclosures, power supplies, single board computers, I/O modules, coprocessor modules and so on. The system integrator would then select components to meet the various system requirements, often from multiple embedded computer vendors. The system would be assembled with the embedded computer components, integrated, troubleshot and tested.
In the past, this approach typically worked well for many systems. But as the capabilities and complexity of each element in the system increased, it was not uncommon for certain nuances of a given product to make it not work correctly with other embedded computer products with which it had never been tested. For example, a graphics coprocessor from one vendor might not work with another vendor’s single board computer–due to the lack of a compatible driver for the graphics device on the selected single board computer, for example.
Rugged System Design
Compatibility and interoperability are just part of the issue. Another challenge, when assembling a system of this type, is to ensure that environmental and ruggedness requirements are achieved. If products are selected from multiple sources–or even from a single vendor but on a “piece-meal” basis–there can be interactions that occur when these items are integrated together that affect the ruggedness of the overall system. For example, if a PMC with a device that runs relatively hot is connected to an SBC in such a way that a “hot” device on the SBC aligns with it, then the result may be a “hot spot” in the chassis. It is possible that the heat dissipation of the system may not be sufficient to handle this “hot spot”–even though all the elements individually have no difficulty meeting the requirements.
If anything, requirements have become even more demanding and more inherently self-contradictory–increasing compute power, for example, while minimizing size and weight–and as component interactions have become more complex, embedded computer vendors who previously only supplied components are now providing complete functional subsystems. This is especially true for UAVs, which are perhaps today’s most demanding application of military computing. That’s because of the stringent constraints placed on characteristics such as heat, weight and size. The design and construction of the subsystem not only leverages those vendors’ intimate understanding of the constituent piece parts and the way in which they interact, it also reduces program risk while freeing the prime contractor to focus on higher level strategic issues and leave the connectivity, integration and testing issues to the embedded computer vendor.
An example of how a subsystem can be specifically engineered for unmanned applications is the GE Fanuc MAGIC1 Rugged Display Computer (Figure 2a), which has been selected for a high-profile UAV program in Europe. The system provides a rugged enclosure with a VPX backplane that can be customized for specific applications. As standard, it comes equipped with a 2.16 GHz Intel Core2 Duo-based processor, connected via 16-lane PCI Express to a graphics processor designed around the NVIDIA G73 dual channel GPU–a GPU that has successfully featured in high-end PC gaming graphics cards (Figure 2b).
3U VPX for Small UAVs
The 3U VPX architecture is particularly well suited to the Small UAV market. The 3U form factor delivers the required compact footprint and low weight, while VPX is capable of delivering substantial computing horsepower. The VPX backplane connector was specifically designed to handle high-frequency serial fabric interconnect technologies such as PCI Express and Serial RapidIO. Pin density relative to VME, for example, is significantly increased while at the same time providing improved signal integrity relative to the CompactPCI architecture in its 3U form.
There’s also an increase in maximum power dissipation available to a single chassis arrangement, which allows for more processing to occur in a single enclosure. While this may seem to be counterintuitive in an unmanned vehicle environment, due to the desire for minimal power dissipation, the reality is that more functionality can be achieved without needing to add a second system enclosure, power supply, heat dissipation system and so on. The overall result is very practical savings in critical elements of small unmanned vehicle payloads.
After system architecture has been selected, the choice of system components is the next critical element of the overall design. VPX may well be a desirable architecture to use, but if the selection of available processing components is severely limited, then the solution may not be viable. However, since it came to market in 2006, VPX has attracted interest and commitment from the majority of players in the military embedded computing marketplace, such that an excellent infrastructure of VPX embedded computer products now exists. This includes not only a range of PowerPC- and Intel-based single board processors, but also multiprocessors, digital signal processors, switches, graphics cards, storage devices and PMCs to provide a broad range of capability.
Video Capture and Transmission
In the case of the UAV program described previously, as with many UAV programs, the requirement existed for a solution that included a video capture and transmission capability. Today, many UAV systems require some form of video data recording, and many require video transmission to a remote ground station. Some UAVs even demand onboard video and image processing technology for real-time, automated, decision-making and analysis. To achieve the required functionality, the MAGIC1 was used as the basic building block, and augmented with a flexible processing module, a PMC card with an integrated FPGA and video I/O capabilities. Each system uses two of these cards to provide the necessary processing capabilities. One card is performing a video capture, MPEG-4 compression and transmission over IP task, so that a remote viewing station can monitor the real-time video from the UAV. A second card is performing onboard custom image processing of another set of video streams for analysis and automated decision-making.
The fact that the identical hardware could be used, with only a firmware/software modification based on the application, was a very desirable solution in a small UAV payload. This provided for fewer hardware components to be managed, maintained and stocked for repairs. It also increased efficiency during system integration as there were fewer variables to consider and expertise with one piece of hardware could be applied to multiple processing elements in the system. In numerous ways, this approach helped to achieve the solution criteria of the UAV market in a way that is only possible with reconfigurable computing elements. Figure 3 shows the same reconfigurable product in two different applications for the UAV program.
The program described provides an excellent example of trends in payloads for UAVs. First, it leverages the growing capability of embedded computer vendors to create complete, functional subsystems, thus providing a valued service to the system integrator. Second, it takes advantage of leading-edge technologies–VPX, the Intel Core2 Duo processor and the NVIDIA N73 GPU–to deliver the significant computing capability demanded by current generations of unmanned vehicle. In particular, VPX in its 3U form factor allows the delivery of a high-performance solution in a small, lightweight package–something that has not previously been possible. Third, this high-performance infrastructure permits the inclusion of sophisticated applications such as video capture and analysis–supporting the growing requirement for aerial vehicles to be not only unmanned, but to be autonomous. The future for unmanned aerial vehicles looks very bright.
GE Fanuc Embedded Systems
Charlottesville, VA.
(800) 368-2738.
[www.gefanucembedded.com].
