By J.R. Wilson
ORLANDO, Fla. - One of the last bastions of increased military spending has been training, especially simulation-based training. But as the Clinton defense budget cuts push ever deeper, that trend too is reversing.
The overall training budget, which stabilized in the mid-1990s after a period of steady growth, now will begin to decline along with other military expenditures, Louis Finch, deputy undersecretary of defense for training, in December told the 19th annual Interservice/Industry Training, Simulation and Education Conference (I/ITSEC) in in Orlando, Fla.
That does not necessarily mean fewer systems, however - just less costly ones.
"There is an interplay between what is possible to be done from the standpoint of technology and a transition in the simulation marketplace in terms of what customers want and can afford," says Rick Maule, vice president-general manager for desktop graphics at Evans & Sutherland of Salt Lake City. "What customers really need is the ability to deploy proficiency training all the way down to the individual unit, if not to the individual. There is a price point that has made that prohibitive before."
What Maule and most other simulation experts see is a new family of trainers, ranging from basic familiarization through stages of complexity until they reach "a level of realism that, to all intents and purposes, puts them in the vehicle and in the operational environment," he says.
But even then, the cost of acquisition, maintenance, and use will dictate dramatic changes from the past and require even greater input and guidance from customers.
"The user has to give us the specifics of what they need to make it a good device for them. We can`t anticipate those," notes Fakespace vice president David Eggleston, whose Menlo Park, Calif., firm is a leader in tabletop pseudo-holographic displays. "We have a good understanding of the physiology as well as the technology. Now we`re pushing that envelope to make it easier for the individual to use the computer to do a job."
The old legacy systems, the big Lockheed Martin CompuScene-based dome flight simulators or those using the big Evans & Sutherland image generators (IGs) soon will fade from the scene - first as new acquisitions (a trend already in place), then in terms of hardware upgrades to existing installations, and finally even to software upgrades. And throughout this process, new, far less expensive systems making full use of the latest microprocessor and software developments will be moving into place.
"Every year people say this is the end of the big systems," notes Bob Grange, advanced graphics product director at Evans & Sutherland. "But there still are a lot of new systems being ordered, especially outside the U.S., and others where the system is old and needs to be upgraded. That work allows us to develop new tools that can flow down to other systems. But there are more programs at the medium- and small-systems level that never existed for visual systems because they were too expensive. You could not have done the CCTT (Close Combat Tactical Trainer) back when simulators cost $10 million each."
As the transition from large to small does progress, many producers long known for their hardware will turn instead to software development. One of those is the Lockheed Martin Information Systems division in Daytona Beach, Fla., which now owns CompuScene (previously GE Aerospace), TopScene (acquired with Vought) and the GT200 (which came with Loral). While TopScene will continue as long as the customer wants it, Lockheed Martin officials say the have no plans for a CompuScene 8 or GT300.
"We won`t build another IG; we`re going into the synthetic environment rather than future hardware," says Steve Buzzard, vice president of business development for Lockheed Martin Information Systems. "We came to a crossroads - put our money into a new high-end IG or go the 3D route and a more commercial market. Our money is going into the software and, at at some point Real 3D (a Lockheed Martin-owned company) will become the hardware generator."
One major reason for this strategy is the Defense Department emphasis on commercial off-the-shelf (COTS) equipment. "Real 3D was our first effort to move from military economies of scale to commercial economies of scale to get the costs down," Buzzard continues. "A larger majority of training requirements can be met by reconfigurable systems - and reconfigurable is largely COTS. We have spent the last two years focusing on internal research so we don`t try to invest in every technology. We can buy displays, for example, and will buy a lot of COTS to keep the price point of simulation down."
In focusing internal research and development on technologies they expect to be key to distributed training and interoperability, Lockheed Martin officials also are striving for common-database, full-system-level interoperability across platforms such as jet aircraft, helicopters, and ground systems.
"Real 3D Pro-1000 is about 90 percent of the capability of CompuScene," notes Real 3D marketing manager Steve Detro. "It actually does a lot more in terms of polygons and pixels at 60 Hz, 16,500 polygons per channel versus 4,000 for CompuScene SG+ for flight simulators. We also do 200 million pixels per second. We don`t claim to be a CompuScene replacement; in some ways, we exceed and in others we don`t do as well, such as lower-fidelity anti-aliasing, 1,024-by-768-pixel resolution, and we don`t do layered weather. But 3D wasn`t designed to be a CompuScene replacement."
Even so, at $20,000 to $50,000 for Real 3D - with commercial discounts to $30,000 per channel - compared to $250,000 to $500,000 for older image generators, that is exactly where the technology is headed, he acknowledges: "As CompuScene fades away, we are striving to make the Pro-2000 higher performance than a next-generation CompuScene would have been. That will be by the first quarter of 1999, with a smaller footprint and equal or lower price (to the Pro-1000).
"Lockheed Martin is taking the CompuScene software, re-porting it to OpenGL and integrating graphics engine technology," Detro continues. "With our Pro-2000, we will be able to take advantage of the legacy technology and be fully compatible as the graphics engine part of that combination."
Evens & Sutherland officials also are moving rapidly into this new environment with their Symphony line, but hold open the possibility of an ESIG 5000 - a decision Maule says depends on customer requirements, although he pledges continued support to legacy systems for the next 10 to 20 years.
"One need is to move at the pace of the market - not just provide the technology, but address the needs at all levels," Maule says. "Not so long ago we could address the high end with fairly proprietary systems because there was no other way to do it. Today you can climb higher and higher with readily available commercial packages. But we also wanted to continue to push the envelope."
But all company leaders looking to address the future of simulation primarily through software and COTS hardware say they realize this approach also dramatically shortens the length of a system`s effective life.
"As you move up through the product line, there is an inertia and a lifetime," Maule explains. "At the low end the lifetime is very short - six to twelve months - and you have to set the bar higher to keep up. As you move up, it becomes more difficult. It would be difficult to build a new Harmony in six months - or for customers to change what they are doing and absorb that."
Part of that evolution also depends on how industry and government customers re-evaluate what level of fidelity they need to meet changing training and simulation requirements. Buzzard, for example, says he sees a continuing need for six degree of freedom dome simulators for pilot training, but not nearly to the degree training specialists have needed them in the past.
Industrywide changes
The entire simulation spectrum is being because designers rely increasingly on fast, cheap processors coming from the likes of Intel Corp. of Santa Clara, Calif. The Pentium microprocessor unlocked that door, the Pentium II pushed it open, and the coming 64-bit Merced chip is likely to open the floodgates. Intel officials already have signed agreements with Microsoft to support the new superchip, but also with such RISC and UNIX pioneers and stalwarts as Hewlett-Packard, Digital Equipment Corp., and even Sun Microsystems.
While Sun leaders vow to continue offering Sparc workstations, they also plan to develop Solaris (their version of UNIX) and, of course, Java for the Merced. All of this will make it difficult for Sparc to continue to compete with Merced-based systems in the future and.
Sun executives also may be forced in the same direction as Lockheed Martin and Evans & Sutherland. Eventually Sun officials may move their venerable work- stations from a proprietary hardware to a software/COTS orientation, including Sun-built Intel Merced-based computers.
While many in the simulation industry are eschewing vertical markets in an attempt to broaden their survivability - sometimes through aggressive acquisition - one fast-growing company often credited with helping drive the move from UNIX to NT is doing just the opposite.
"We have been an independent company. We haven`t done any acquisition and aren`t interested in doing much, Robert Thurber, executive vice president of Intergraph`s simulation systems division in Huntsville, Ala. "Rather, we`re looking at verticalizing ourselves. We`re trying to develop the right kind of relationships with good partners that don`t require investment. I see partnerships being formed, but I think they`ll be longer term rather than shorter."
Price - with power - is a significant factor in the hardware and software sides of the battle between UNIX and NT. An NT machine can access far more software at far less cost - even NT versions of existing UNIX software - but UNIX software often has more power and reliability.
The same is true on the hardware side - a top-of-the-line Intergraph TDZ 2000 3D graphics workstation with dual 300 MHz Pentium II processors and half a gigabyte of RAM is about half the cost of a comparable SGI Octane system. On the other hand, users can configure the Octane with more than 16 gigabytes of RAM and multiple processors with little degradation of power in the additional chips.
Pushed by Intergraph and other workstation competitors, Silicon Graphics executives also have cut the price of their low-end O2 series, which now starts at $5,902. Intergraph, however, offers NT workstations for one-third the money and comparable - if not superior - performance.
Thurber also is quick to dismiss arguments that NT cannot handle the computing needs of high end simulation: "Windows NT satisfies the real need for training simulations - you don`t need a high-end real-time OS to do that," he says. "You do need it for hardware-in-the-loop simulations, where you have to handle IR seekers, for example - but not for training or mission rehearsal or virtual prototyping. The current NT limitations make it difficult for perhaps 5 percent of this market, but for the other 95 percent, NT is just fine.
Hardware for new simulators is moving toward commercial off-the-shelf solutions, rather than the proprietary hardware that was the rule for previous generations of simulators. The Evans & Sutherland Rhythm product, pictured above, is a single-channel, single-board image generator combining proprietary REALimage graphics with PCI bus, video controls, and external genlock features.
