On a quiet street just off of Nashville's historic Music Row, a dedicated team of more than 100 researchers are developing software systems that may very well revolutionise the modern world. From cutting edge cyber-security tech designed to defend America's most critical networks to computer-learning algorithms that could all but eliminate airline crashes, Vanderbilt's Institute for Software Integrated Systems (ISIS) is making our Internet of Things safer, better, stronger, and more resilient.
I recently sat down with Professor Doug Schmidt, Associate Chair of the Computer Science and Engineering Program at Vanderbilt University, to learn more about what goes on at this secretive research centre. You won't believe what they've got in store.
Gizmodo: Can you tell us a bit about the projects you're currently involved with?
Doug Schmidt: We do a lot of work with various government agencies, like DARPA and the military service labs. We do a lot of work at places like the National Institutes of Health and the National Science Foundation. We also do work with industry, like Microsoft, Siemens, or General Electric, and with companies like Lockheed Martin, Boeing, and so on.
We have about a dozen different projects right now. One of the things that we've been doing has been focusing on trying to make it less expensive for the government to build large-scale defence systems. We're doing something called "open systems," trying to build systems that use standards-based off-the-shelf parts so that people don't have to spend a fortune trying to build them all in a proprietary way.
We've been focusing on something called the future airborne capability environment, which is basically a common operating platform, composed of standards, reference notation, performance tools, and so on. We can use that to reduce the cost of acquiring software for military avionics systems — things like Joint Strike Fighter, the F18, and so on.
We've also been spending a lot of our time working with DARPA to come up with a way to revolutionise how military vehicles are built. So, rather than going in and having people go on these things manually with screwdrivers and blow torches or bending metal by hand, we're building model-based tools that allow designers, architects, and systems engineers to essentially make the effective equivalent of a CAD diagram. Then we can analyse those diagrams for various properties electronically and computationally.
Once people are comfortable and happy with how it looks in the simulator, then they can go ahead and actually develop tools that will fabricate the result. It's something kind of akin to 3D printing if you will. Rather than doing everything manually, you can use computer-aided design tools to do the composition, the synthesis, and the analysis at the design level. Then, when you're happy, you push a button and out come specs that are fed into robots to make the actual thing you're trying to build.
Giz: That's fascinating. Now are the design aspects off-the-shelf components as well? Or have you guys had to develop custom software?
DS: What we've actually done for, gosh, 10 to 15 years or so, is building a suite of tools that make it easier to design software. It's called the generic modelling environment — GME — and what we do is take the infrastructure and the tools that we've built in-house, and we've customised them on something called the advanced vehicle make program, and we customise them from the domain of designing vehicles.
The Adaptive Vehicle Make research program could cut vehicle design times by 80 per cent
So we didn't start from scratch. We started from an infrastructure we already had, and we've customised it and made it so that the main experts — subject matter experts, systems engineers, architects, and so on — can then use tools that work the way they are comfortable working, and then they're able to use that stuff to do the analysis, the synthesis of the design.
Also, the military tends to spend a lot of money on custom radios, and custom communication systems that are chronically late, over-budget, or don't work. So one of the things we've been doing here at ISIS is that we're taking commodity software and hardware like Dell products — tablets, smartphones, and things like that — and we're combining open-source, open-system software like Android, then we're adding secret sauce in a few places having to do with security, making a mobile network as opposed to fixed cellular networks, making things be more reliable, making them be so that you can't compromise them through various kinds of cyberattacks, making them work applications that are kind of tailored for soldiers, tracking themselves and their opponents.
The AMMO system in action
The cool part about it is that we're able to give people essentially the experience you'd have as a commercial user in the USA with your smartphone, except we're able to do these things in far-off lands where there's no fixed cellular network, where it's much more ad hoc, and where the radio network's almost weirder and more specific to the DOD, we're able to give the soldiers basically the same experience, or even a better experience really, than what they would have back home, while they're out doing their mission.
Giz: How computationally intensive are these programs? Can they run on, say, desktop computers or do you need specialised computational hardware in order to use them?
DS: For doing the things that we do — the equivalent of CAD tools where you're laying things out and connecting them together — those things can run on laptops or desktops, no problem. When you get to the later phases where you're trying to do analysis of the models in order to check them for certain properties, that's where you need the big iron.
I think an equivalent analogy would be people who do animation. You would have artists who would be working on desktops with accelerated graphics cards on them, and they get kind of a scene played out. Then, when they're done, they turn that data over to a render farm, which is a big cloud or data center where they do all the compositing and rendering of the final product.
Giz: And what instigated the development of that program? Was there a specific need for it, or was there just more of an overarching general need for the technology?
DS: Not unlike the first example I gave you where we're developing open, systems-based approaches for building defence acquisition systems, there are two motivations: number one, it takes way too long to build systems in the conventional way and, number two, it's way too expensive to build systems the conventional way.
The government and the defence department are... reducing the amounts they're spending, or, more accurately, they're reducing the rates of growth of the money that they're spending. They're starting to look for new ways of doing things cheaper/faster/better, and so, in both cases, it was a drive, a realisation, it was taking too long to build military vehicles, it was too expensive to build these vehicles, and everything that was done was done in a customised, proprietary way. That was not really where the government wants to be.
They want to be able to have something that's equivalent to the commercial car manufacture market, where it's relatively inexpensive to get relatively good quality stuff, there are a lot more products, and it's not done in a custom way. You want them to cost $US30,000, not $US30 million.
Giz: Right. And what sort of cost and times savings are you guys looking at? Basically, how much more effective is it than the conventional method?
DS: It's orders of magnitudes faster. Essentially what we're doing here, or the way to think about this, is what people sometimes call mass customisation. In the commercial world of — let's just use automobiles as an example — you, the end user, get to customise a handful of things like the colour and the interior, things like that. Everything else is commoditized. Even though things are relatively expensive in the grand scheme of things, it's a lot of pre-assembled parts that are taken off the shelf and get plugged together.
A GPS smartphone tracking device for sniper fire location
You look at the military space... literally every piece for every aeroplane is different, for various reasons having to do with stealth technology. It's ridiculously expensive. So we're getting orders of magnitude improvement in cost, while still being able to deliver things that are specific for the more boutique needs of the defence department. There are things that the military needs but that they would prefer to be able to do with off-the-shelf parts that they just harden or customise in a few ways. They want to make the customisation process similar and rigorous but as automated as possible.
Giz: And have any of these technologies reached the consumer market yet?
DS: A lot of the stuff... I think the closest thing in the consumer market is the buzz around stuff like 3D printing, not necessarily using it for building military vehicles but people are starting to use this to do, I don't want to say mass customisation, because it's still fairly expensive, to do stuff with 3D printers, but you can do stuff a lot... you can do rapid prototyping, development of things in a very unique, larger and, of course, cheaper scale than if you had to do them manually.
In the Geoffrey Moore, Crossing the Chasm scheme of things we're sort of at the "early adopter" phase. We haven't crossed over into the mainstream, we're still crossing that chasm. But I think as more people adopt this stuff the prices will come down, the comfort using the technology will increase, things will be much better.
Images: Vanderbilt University