For several decades, the military has been using large-scale client-server systems to build networked environments where soldiers can train in simulated battle conditions. Now the military is looking at peer-to-peer technology as a way to build these simulations without a vulnerable central server. Michael Macedonia, the Chief Scientist and Technical Director for the U.S. Army Simulation, Training, and Instrumentation Command (STRICOM), talks with O'Reilly editor Richard Koman about how the military simulates battle, how peer-to-peer technology could change that, and the advances that have made a $69 flight simulator program as valuable as the multimillion dollar systems of a few years ago. Michael Macedonia will talk at the O'Reilly's Peer-to-Peer and Web Services Conference, September 18-21, in Washington, DC.
Koman: What's the military's interest in peer-to-peer technology?
Macedonia: We're trying to develop a model here of simulation that essentially provides a very powerful computing capability to every soldier in the Army. We foresee over the next three to five years a need for highly networked, very powerful mobile computers. We have a buzzword here. We call it "simulation on demand," so that when you need to learn something, when you need to experience something, when you need to plan something, you can instantly get it.
We have a number of programs, for instance, the Close Combat Tactical Trainer, which is a vehicle simulator that's networked together, and you can go play combat with 40 other crews in a building.
Koman: Can you walk us through that scenario?
Macedonia: The way the Close Combat Tactical Trainer works is that you have modules that represent Bradley or Abrams fighting vehicles, and typically there are about 40 of these modules, and they're configured based on what the makeup of the units are at that location. So at Fort Hood, Texas, they have about 25 tank modules and about 15 Bradley modules in one building. And these modules, if you go inside them, look identical to the combat vehicle itself.
Koman: They're full-scale models of the real thing?
Macedonia: Well, in the crew compartments. I mean, if you stand outside, they're definitely not full size because there's no gun on them or anything. They're simulators. The crew gets in the simulator and they look out through the optics or the fire control and see a virtual environment. We've got databases that represent the National Training Center, which is in the Mojave Desert, for example. We have a central European one. We have a Kosovo database. We have a Fort Hood database. They could actually go virtually fight the back forty at Fort Hood--cow patch, creek, whatever. So each of these forty modules is networked to each other over an Ethernet, so essentially since it's a peer-to-peer environment equal in computing capacity.
But it's rather limited. Now you can tie these different sites together and create virtually larger units. What we'd like eventually is to have everybody able to carry that environment with them. So they could be wearing a head-mounted display (HMD) and have that synthetic environment projected into the head-mounted display. So basically, folks could go fight an augmented reality and actually fight against synthetic characters that are projected into their HMD. They would have the same physical challenge of running around buildings, but then they would see virtual enemies, good guys, or virtual neutrals, like virtual police.
Koman: And the individuals are linked together?
Macedonia: All linked together. What we're trying to do right now is sort of baby-step and we're looking at different communications technologies to be able to do that. One idea is using an 802.11 LAN with everybody. We also have some digital radios that we've been messing around with.
Koman: So what would these simulations look like before peer-to-peer?
Macedonia: Well, we have an example of it over at the Navy Research Labs (NRL)and at Columbia University, and there are also huge research labs out in Malibu that have been working on augmented reality systems. The Columbia system was shown at SIGGraph this year. They basically take a Dell Inspiron laptop with an Invidia graphics chip in it and strap it to a guy's back, and they tie it to a head-mounted display and a high-precision tracker. So, for example, if you looked at different buildings or different booths and stuff like that, it would automatically tell the wearer of the system what they were looking at. So it was great for spatial recognition. The Navy Research Lab has a very high-precision, differential GPS on it, and it's good down to, I believe, one centimeter.
Koman: So it's displaying your GPS location?
 |
|||
|
Michael Macedonia is a featured speaker at the O'Reilly Conference on Peer to Peer and Web Services, Nov. 5-8 in Washington, D.C. Related Articles:
Next Step for P2P? Open Services Convergence of Peer and Web Services O'Reilly P2P and Web Services Conference: Program Chair's Best Bets | |||
| |||
Macedonia: Yes, but it's really good because you can look at a window and it will tell you who's working on the other side of that window. If you're walking around the buildings at NRL with the system on, you can look up at a building and it'll tell you what the room number is and who's in that room. I mean, it's very spooky when you wear this thing.
Koman: So is everybody wearing a GPS so we know who's on the other side of the window?
Macedonia: Well, for example, you could annotate a window and say, "There was a sniper here yesterday." If the next day somebody's wearing the system, he would know, "Watch out, there was a sniper here yesterday," or, "This is where we left the ammunition," and so on. So that's one step. In that application, you're using it as a navigation tool. We're pushing it a little further as we have the Institute for Creative Technologies out at UCF working on it. Our interest is bringing in these synthetic characters, or synthespians, and letting them play in your world. I mean, frankly, if I can have 3-D graphics superimposed onto my display, then I could put synthetic actors, as well as virtually represent real people who are actually someplace else.
|
Koman: So let's say you have a program of a terrorist situation, and you've got snipers in the windows....
Macedonia: We have 400 actors on our payroll at STRICOM. Actors are wonderful but they get tired, they whine a lot, and they're very expensive. Just like real actors in Hollywood, they're always wondering what the motivation is. One of the things that we're interested in is adding synthetic characters that represent police, the Red Cross, kids, and so on; essentially things that are usually in the real world that don't get represented a lot in military training. Creative Technologies is trying to create synthetic characters that have real behaviors that can talk to you and interact with you with emotions and behavior in a rather restricted domain. So imagine if I've got all these people hooked up to these augmented reality systems, and I'm able to inject all these synthetic actors into it.
Now one of the major theme park companies is very interested. I will not reveal their name though, but did you notice that we're in Orlando?
Koman: I bet they would be interested. It sounds an awful lot like the Holodeck.
Macedonia: But seriously, this is serious research, and there are a number of other organizations and outfits working on similar ideas. For example, Canon has been working on this technology. So we're at the phase where a lot of things have started to change to enable this to come out of the lab. Number one is portable graphics-processing capability. Right now you can actually do 3-D graphics on a Compaq Ipaq. If you just follow the normal curve, graphics computing power doubles every 12 months. It's actually faster than Moore's Law, which is every 18 months. So you're seeing mobile processors rapidly advance. Then you have just gobs of memory. Memory is no longer an issue.
Once we get into the high-bandwidth wireless issues, whether it's 802.11, or it's 3G or 4G, we can actually have huge, peer-to-peer mobile computing environments, because from a military context, having a centralized server is a point of failure, a critical failure node. You don't want to put all your data on one server because once you take that server out, then you've got a lot of blind people with a lot of useless electronics.
Koman: Right. The theory of the way the Internet itself is built.
Macedonia: Then the issue becomes, I have a lot of devices that in a sense become servers themselves. I mean that's the whole idea behind P2P.
Koman: Right. They're devices and servers at the same time.
Macedonia: So in fact, we're starting to demonstrate that on some of what we call semi-automated forces staff, which is sort of the AI bad guys we use in our simulations. We used to run these on small supercomputers; now we're running them on PCs with Linux.
Simulations don't lend themselves to a centralized server because from a technology standpoint, the server becomes the bottleneck, and for years I've been involved with trying to come up with schemes using multicast communications and stuff like that to enable peer computing. Multicast still has a long way to go, but in a wireless environment you essentially have a broadcast environment, and you don't have all the issues of routing bandwidth, like you do with like things like Gnutella and FreeNet.
Koman: Can you move from this fairly sophisticated simulation process of peer-to-peer communication between modules to a life combat situation?
Macedonia: Yes, in fact, we've got tons of research that shows that people's performance is substantially improved by training in simulation. I mean oodles of data. It's the most common question I get, and I say yes, we can make people really good in simulation.
Koman: But in combat itself, you've got soldiers with head-mounted displays, perhaps. Can you get live data in a real combat situation, shared in a peer-to-peer fashion?
Macedonia: That's the Holy Grail. We have lots of data that shows that people in simulation do extraordinarily well in the live environment. Now, the thing is, they've always been two separate events. Let me give you an example. In the '80s, we built a gunnery trainer called UCOFT, Conduct of Fire Trainer. With UCOFT you can fire tank rounds day and night because they're virtual tank rounds; they don't cost you any money so you just fire virtual tank rounds all day long. We took those soldiers out and we put them in a real tank together after using UCOFT, and these folks went and beat everybody in tank gunnery imaginable. Everybody said, "How did these soldiers get so good?" Well, because they spent literally days and weeks inside this gunnery trainer perfecting their gunnery skills, as opposed to the typical training where the soldiers go out and they fire a couple rounds a day when they're in training.
If you look at the future, your training system suddenly becomes your mission-planning system. Before I go attack that hill, I'm going to run a simulation of it with my squad over the next 10 minutes, and we're just going to simulate it and we're going to do it virtually while we're waiting here for orders to move out. Before you do anything you go out and you run a simulation.
|
Koman: So, at a strategic level there's war gaming, where you're planning out the whole thing, but what you're talking about is sort of 10 minutes before the battle, let's just plot out what we think we're going to do and what the likely outcome is.
Macedonia: That's right. And go virtually fight it.
Koman: Well, that's pretty amazing.
Macedonia: That's called "simulation on demand."
Koman: So what's the input from field commanders?
Macedonia: They love the idea. Essentially the folks who'll have the edge anywhere are the folks that not only have the best talent, but they have the best talent that have worked together and experienced whatever it is that they do in training and preparation, and to the highest state possible. Desert Storm was a good example of that. We sent 500,000 people and only 250 Americans died. We won in four days and I forget how many Iraqis died; most were poorly trained draftees from within Iraq.
Now I'm going to tell you a true story. Two years ago, this Navy pilot shows up at Pensacola Naval Air Station. That's where they teach the new recruits how to fly. And actually he was an ensign; a brand new Navy aviator. So he goes out and they take them all on a check ride, and he maxes every check ride. Well, they hadn't had that happen in years, so they said, "How come you're so good? How long have you been flying?"
He says "I've never flown a plane in my life until I got here."
And they said, "Come on. What's your secret?"
"Microsoft Flight Simulator." He had hacked a database and built a database of Pensacola, and so before he goes out on his check ride, he goes and flies in Microsoft Flight Simulator. It's only $69, and the new version of Flight Simulator is really, really good. Now I get all these debates stating that the dynamics aren't right. I say, "B.S." The program manager behind Microsoft Flight Simulator program was a program manager at Boeing. The dynamics in these systems are alright. There's a GAO report that says, "Microsoft Flight Simulator is appropriate for training people, particularly in basic instruction in aircraft."
Koman: So sitting in front of a piece of software on a laptop is just as good as being in a million-dollar flight simulator or a 35 million-dollar flight simulator?
Macedonia: For certain things, and they're primarily what we call "command and control skills" or "situational awareness skills," which is navigating. It turns out that one of the big reasons pilots get confused is navigation. It turns out Microsoft Flight Simulator is actually very good for those type of skills, and GAO wrote a big report up and the FAA approved Microsoft Flight Simulator for use in flight schools. So we've gone from $35 million to a PC game, and in a couple of years we'll be able to have that level of software on very portable devices that we can walk around with all the time.
Koman: I could imagine that the military could have a contract with Microsoft to build a database of aircraft carriers and military landing strips?
Macedonia: They're not interested. We've tried that. They don't want to get into custom work for the military. But the bottom line is that you can now have very powerful simulations on very powerful but small computers tied to the Net, and have people interact. One of the big things that's coming out today, or for the last two years, has been EverQuest, which is a 3-D virtual reality game. There are other ones called Asheron's Call and Altima Online.
Basically they're virtual environments or simulations that you get immersed into online. I think this is the real revolution, because for Microsoft the persistence is in the database that they give you on your individual PC with the CD-ROM. These games connect to servers, so it's not pure peer-to-peer, but it's interesting that people are already hacking them so that they run servers on their local machines. But basically they're persistent virtual worlds. That is when you come back to your PC and get online, the house that you built virtually still exists out there in the world, and all the other players see the house that you built, or that the person you killed is still dead.
|
|
|
|||
|
Michael Macedonia is a featured speaker at the O'Reilly Conference on Peer to Peer and Web Services, Sept. 18th-21 in Washingtonl, D.C. Related Articles:
Next Step for P2P? Open Services Convergence of Peer and Web Services O'Reilly P2P and Web Services Conference: Program Chair's Best Bets | |||
| |||
Koman: So you can imagine a military simulation of a three-week-long engagement?
Macedonia: Right.
Koman: And you mentioned at early in our conversation that some theme parks were interested in some of the work you're doing. I'm wondering if military technology is driving the entertainment industry, or vice versa, or whether there's sort of a 50-50.
Macedonia: I don't know. That's a good question. Frankly, to a large extent it's the commercial world driving what we're doing. I mean Microsoft will spend 2 billion dollars on the XBox, and I think 500 million of that is on marketing. The entire Army budget for research and development this year is $1.6 billion. So Microsoft is spending more money on a game console than the Army is spending on basic and applied research. So we're in a dilemma here. We can't outspend Microsoft.
So in a sense the advancements in the technology is mostly coming from the commercial sector. For example, in processor capabilities, in storage, in memory, in computer graphics--that's all in the commercial domain. Now in AI it's a different story. We have certain niches that essentially are complementary. Now the game community is becoming interested in artificial intelligence. We've been funding a lot of programs and have been a leader in that area for a long time, for a variety of reasons. But they're niche technologies related to simulation and training.
Koman: I read about the Intel Pentium 4 yesterday, coming out with a 2GHz processor next week, but the architecture is designed to scale out to 10GHz.
Macedonia: I don't know when they'll hit 10GHz. I presume probably in the 2006 time frame, or 2007. Don't ask the military to go develop a custom processor anymore; we're just going to watch Intel punch out this processor.
Koman: Back to peer-to-peer--does it seem ironic at all that you're applying some of the concepts that come from some of these services that are fairly subversive, at least as far as the recording industry is concerned. You know, Napster-style ideas applied to military technology.
Macedonia: I don't think it's subversive. The only interesting thing about Napster was that they came up with a really good scheme for sharing music. I mean this subversive thing is just in terms of the way that the RIAA or the MPAA looks at this technology and sees it as a threat to IP rights.
Koman: One does have a sense that the military is a top-down command sort of enterprise, and you're talking about a decentralized way of communicating information and of creating simulations.
Macedonia: Well, in a sense, but this picture of this monolithic hierarchy in the military is about the biggest mistake most people make. Because the reality is that in terms of information technology, it's very diffused within the military. In fact, from one standpoint, you'd say the Navy Marine Corps Internet Project.
One of the problems that the military's had for the last couple of years has been trying to consolidate MIS resources because everybody out there has been able to go out and buy their own computers and networks and set up shops all around. The Army's now going to something called Army Knowledge Online, and they're centralizing the funding for computers, networking, and MIS management. They're not necessarily centralizing the operation, I think that'll be the next step. They're trying to come up with an AOL-like model for the Army, for computing. But the reality is that we're part of this global infrastructure. The same networks that civilians use to send Internet traffic, the military is buying time off of too.
So we're using the same satellites, the same technologies, and in general the same networks. We lease stuff from the major telcos. The history of communications has always been about control. But who really has control over all that fiber in the ground and more fiber eventually to go into the ground, with wireless, etc.? Every MIS shop in the commercial world as well as in the military world is trying to establish some level of control. But how much control can you have when everybody is basically operating five or six devices?
Koman: Right. Each individual has a substantial amount of computing power at their personal control.
Macedonia: Now, AOL probably has more control than anybody else. But I think the key question for me, and what we're trying to answer at STRICOM is not who controls information, but how to use the technology most appropriately to ensure that our soldiers have the best training available. And that's the bottom line for us.
Koman: Right. So, bottom line, applying peer-to-peer to military simulation is improving the quality of the simulation.
Macedonia: And improving training.
Koman: And saving lives?
Macedonia: And saving lives. Exactly.
Register now for O'Reilly's Peer-to-Peer and Web Services Conference and hear Michael Macedonia's talk about Network-Centric Warfare.
Richard Koman is a freelancer writer and editor based in Sonoma County, California. He works on SiliconValleyWatcher, ZDNet blogs, and is a regular contributor to the O'Reilly Network.
Copyright © 2007 O'Reilly Media, Inc.
