Page 1 of 1
There are perhaps as many definitions of situational awareness as there are situations to be aware of. A consensus exists, though, that in its simplest form, situational awareness is about knowing what’s happening so that an informed choice about future actions can be made. As such, it’s not surprising that, as a term, it was first used by the military—specifically, U.S. Air Force aircrew in Korea and Vietnam. It’s equally unsurprising that providing improved situational awareness to deployed forces is the focus of much investment on the part of the military.
What drives situational awareness is knowledge. That knowledge derives from information—and what drives that information is data. As Clint Eastwood famously pointed out in For A Few Dollars More, “a man’s life in these parts can depend on the merest scrap of information.” He might have been talking about the military’s requirement for superior situational awareness.
The military has never been positioned to gather more data than it can today. It can be gathered from almost any source: satellite, unmanned aerial vehicle (UAV) (Figure 1), traditional reconnaissance, databases, vehicles—even individual troops. It can come from cameras, sensors, the network. It can take an enormous variety of forms—video, graphics, infrared, sonar, radar, lidar (ladar).
Four Computing Challenges
Collecting that data is a significant computing challenge. That said, it is but one of four computing challenges in situational awareness applications. Only a fraction of the data collected is valuable at any one point in time: delivering to the user huge amounts of raw, unanalyzed data will serve only to overwhelm and confuse. Situational awareness requires that the information presented be easy to assimilate and easy to understand—complex data must be easy to make sense of “at a glance”—if it is to form the basis of rapid action. The second computing challenge, therefore, is to winnow out the irrelevant data. Paradoxically, as significant an amount of computing power is deployed in discarding data as was expended in acquiring it in the first place.
For the information to become knowledge, it must, however, be complete. Information about the location of enemy vehicles, for example, is incomplete without information about the capabilities of those vehicles, the intervening terrain, or the position of friendly troops. Information about a target is inadequate if nighttime conditions preclude a detailed view of it. The third computing challenge is to fuse together different sources of data and to present it as meaningful and actionable information. And, finally, the fourth computing challenge is the requirement to take the gathered, processed and fused information and to display it on-screen in close to real time. Information that is only a second out of date may be valueless—and possibly even disastrous—information.
Situational awareness is, by definition, an application that will, as a result of complex and demanding processing, output sophisticated graphics—whether as video, still images, symbols, icons or animations. From data acquisition to image output, it is highly computationally intensive and requires a computing architecture that offers the requisite bandwidth and raw processor power. Moreover, the optimum computing architecture will provide a natural platform for graphics-oriented applications because it will provide the potential for seamless integration with state-of-the-art graphics hardware. Increasingly, it seems that VPX is that architecture.
VPX Delivers the Goods
First and foremost, the VPX architecture is designed specifically to leverage the high performance that is inherent in serial switched fabrics such as PCI Express, Serial RapidIO, Gigabit Ethernet and InfiniBand, and their capacity to support the very high data transfer speeds required in networking, data storage (Serial SATA) and digital video (such as DVI, LVDS). Applications such as situational awareness require significant backplane streaming I/O, connectivity and bandwidth: VPX provides not only substantially more I/O pins, but it also enables bandwidths of up to 30 Gbytes/s—enabling more data from more sources to be captured than is possible with the more limited pin count of VME.
This significantly increased I/O derives from VPX’s implementation of the Tyco MultiGig RT2 connector (Figure 2), which is rated for signals up to 6.25 Gbits/s
—around six times the capacity of the VME connector. Historically, connectors have incorporated pins arranged such that the vertical and horizontal pitch positions of the contacts are symmetric. It’s an approach that works well for single-ended signals. However, the differential signals that characterize switched fabrics operate better over pairs of pins that are close to each other, but that are as distant as possible from other pairs. Add to this the increased requirement for good grounding that arises as a result of much faster signaling speeds, and the need for a new connector technology becomes apparent. A 6U VPX board features six 16-column 7-row RT2 connectors and one 8-column 7-row RT2 connector, while a 3U board features two 16-column 7-row RT2 connectors and one 8-column 7-row RT2 connector.
Another, and not inconsiderable, benefit of the VPX architecture is that switched serial interconnects provide the foundation for highly robust, fault-tolerant architectures that can continue to operate in the event of the failure of an individual board. This contrasts with the parallel bus architecture of VME (or CompactPCI) in which failure of any single board will cause a total system failure. Given the mission-critical nature of situational awareness applications, this element of high reliability assumes great significance.
Support for Powerful CPUs
VPX also provides for the support of powerful processors, enabling both the processing required to turn the mass of data into a much smaller amount of information, and also to fuse that data together in order for it to represent knowledge. Where the VME specification allowed for no more than 30 watts of power dissipation per slot, VPX allows for significantly more: 115 watts at 5 volts, and up to 768 watts at 48 volts.
It does this through its use of dedicated power connectors in P0—as opposed to via signal pins in P1 and P2 as is the case with VME and, for that matter, CompactPCI. The implication is that VPX can accommodate the power dissipation of leading-edge processor silicon. For example, Intel’s Core 2 Duo/Merom processor, designed specifically for low-power environments such as mobile computing, still dissipates 35W, and its 45-nm Montevina successor will allegedly dissipate 29W. Meanwhile, Freescale’s 8641D dual core processor dissipates somewhat less, with around 19W at comparable processing speed. But more importantly, VPX accommodates the increasing functional density of today’s highly integrated single board computers.
The combination of high-bandwidth I/O and powerful processing—which includes not only traditional general-purpose processing, but also multiprocessing, digital signal processing and FPGA-based processing—that characterize the VPX architecture make a formidable contribution to its suitability as the basis for a compute-intensive, data-intensive application like situational awareness. It provides for the acquisition of substantial amounts of input data from a variety of sources, and for processing that data. In fact, it goes beyond the requirements of today’s applications and provides the possibility of enhancing them.
Take digital mapping, for example—a key element of situational awareness applications. Many of today’s digital mapping applications feature 2D, rather than 3D, presentation. One of the factors contributing to this is that the data points from which the terrain data is derived are sufficiently far apart that, when rendered in 3D, the resulting image appears as a number of jagged edges, rather than the smooth contours that are the reality. Sophisticated software exists that would allow this image to be smoothed in multiple directions—but the processing overhead invoked is considerable. The support provided by VPX for greater bandwidth and more powerful processing would allow this processing to take place—thus improving the quality of the map presented on screen.
On-Screen Data Challenge
Historically it’s been the presentation of on-screen data that has most challenged military applications. Here again, though, VPX brings important benefits when PCI Express is implemented as the primary serial switched fabric—as it is on the GE Fanuc Embedded Systems range of VPX single board computers.
The attraction of PCI Express is that it is the native interconnect technology for desktop computing, and thus provides simple and cost-effective access to the enormous number of devices and technologies designed for the PC environment. This is especially true of PC gaming technology, as characterized by products from companies such as NVIDIA—technology that promises to revolutionize not only situational awareness applications, but also other military applications such as simulation and embedded training.
The VPX-based GE Fanuc Embedded Systems MAGIC1 Rugged Display Computer (Figure 3), for example, provides a 16-lane PCI Express interconnect between an Intel T2500 Core Duo CPU running at 2.0 GHz and an NVIDIA G73 GPU in order to deliver optimum throughput and performance. It also features up to 64 Gbytes of SATA-connected storage, and a Gbit Ethernet interface. The SBC340 single board computer at the heart of the MAGIC1 is specifically designed to take maximum advantage of the silicon potential of both the Intel chipset and the NVIDIA GPU, the latter being the ideal platform for providing the real-time graphics processing—smoothing, fusing, overlaying and so on—required by the application.
Such a display processor might well find itself deployed in a fighting vehicle, where an element of the situational awareness requirement is to deliver to the crew a 360 degree view of the environment in which they find themselves (Figure 4). This view is generated by an array of distributed aperture video sensors (both TV and IR), which must operate in real time in order to be operationally effective. Previous processing architectures have not had the “bus power” to handle multiple, real-time video signals; VPX, with its associated PCI Express bus, now offers this capability.
Multiple Sensor Views
Like all elements of situational awareness, this multiple sensor view is of maximum value only when it has been “processed” in order to highlight items of interest to the overworked “operator.” In order to meet this requirement, Distributed Aperture Sensor (DAS) systems have both a detection and classification element (data extraction) and an image stitching, aligning and formatting element (visualization). The former is a highly computationally intensive task for which the VPX CPU is ideally suited, while the latter can most effectively be accomplished in the VPX GPU.
The 3U VPX form-factor overcomes many of the inherent disadvantages of VME in its 3U guise, making it ideal for deployment in space-constrained, weight-constrained environments. The input from image sensors/cameras around the vehicle will be correlated and stitched together into a cohesive, panoramic whole. If necessary, infrared images will be overlaid, allowing detail to be discerned despite the lack of light or, for example, because of the presence of smoke or mist. Overlaid will be terrain and mapping data that, together with the acquired visual data, will provide a complete, fused view of any possible threats—before the hatch is opened.
There can be little doubt that the VME architecture will continue to be at the heart of demanding military embedded computing applications for many years to come. However, it looks likely that, for demanding graphics applications such as situational awareness, VPX brings with it a set of characteristics—substantial I/O bandwidth, ability to leverage leading-edge processor- and board-level silicon, native access to state-of-the-art graphics processing and inherent fault tolerance—that will see it become the architecture of choice.
GE Fanuc Embedded Systems