Sensing and Sensibility

The future of sensing for the US defense industry, according to the Head of Maritime Sensing, Michael J. Wardlaw at the Office of Naval Research (ONR), depends on two elements: our ability to successfully extract useful and relevant information from an ever-increasing deluge of data, and our willingness to rethink our approach to platform-building. "When I build a sensor," he says, "it's necessarily built with the intention of going on some thing, some platform - a ship, a boat, an aircraft carrier. It's nearly impossible to develop naval-related technology without having a target platform firmly in mind."

For Wardlaw, who gave the keynote presentation at the SPIE Defense and Commercial Sensing's Ocean Sensing and Monitoring Conference in April, some contemporary technology trends seem quite clear: technology is "aging out" at an increasingly fast rate; at the same time, it takes longer to build and deploy our increasingly complex platforms and systems. Recognizing that these trends will inherently create challenges for the Navy suggests that a fundamental rethinking of our technology development strategies is appropriate. To begin with, Wardlaw points out, most sensors tend not to be all that smart. "They're typically built to simply collect data," he notes, "in much the same way as laboratory instruments. It's an open-ended approach, allowing users to swallow up every scrap of data possible. This is usually done because it's nearly impossible to know what data will actually be needed in the future. Of course, sensor designers are probably correct in that assessment, but there are some significant consequences to taking this approach." The necessity, he continues, of incorporating deep-learning machines into sensors as well as building dynamic and agile platform systems to support those sensors seems increasingly clear.

Consider, for example, that most sensors currently being designed and built almost always start with a platform in mind. "That single fact turns out to be a huge design and development constraint," says Wardlaw. "Naval platforms, as we develop them now, are intended to survive for decades. So when I conjure up some wacky idea for some bizarre sensor and I want to put on one of our platforms...well, there isn't much appetite for anything disruptive on a platform that needs to be tactically relevant for decades and available in a moment's notice. That's what I call ‘a natural disconnect.'"

But clear challenges also provide opportunities for clear solutions. In Wardlaw's workspace, he's busy developing what he calls "the splinter force," a playground for disruptive technologies. It's a somewhat disruptive concept in its own right, and makes everything else he's up to all the more future-focused. "Imagine what the implications are if - without disrupting the traditional acquisition programs - we have the opportunity of changing just one of the major constraints," he posits. "Instead of designing a platform to last for decades, imagine if the platform only needed to survive a mission or two - only needed to be utilized for a couple of weeks - and then you could chop it up and make something else out of it. I'm talking about platforms that can serve an immediate application, can support the specific sensor that's needed at that particular point in space and time, and then can be disposed of or recycled."

A recent ONR project explored innovative efficiencies in re-charging autonomous undersea vehicles while they were collecting and disseminating information, kind of like an underwater 7-11 for robots, as Wardlaw described it at the time. It turns out that the resources that are particularly valuable in the undersea environment, he says, are power, communications, and the ability to dock; those were the three fundamental components explored in that program. "However, if you were to question each and every individual naval officer on how much power, how much comms, and how much docking they needed, every single one of them would likely give you a different answer," says Wardlaw. "Therefore, it seemed rational to me not to get hung up on what the answer was, but to figure out what the process was. I could focus on those three key components and then come up with a process that would be able to dynamically allocate those components as resources."

Wardlaw's current focus remains an extrapolation of that same fundamental concept, he says, referencing Timothy P. Grayson's DCS plenary presentation: "I have a sonar here, I have a camera there, I have different pieces in this puzzle. It's like DARPA's mosaic process - we're talking about the same thing. It's the plug-and-play notion of what it takes in order to be able to focus on developing the assets, the resources, and using them interchangeably. Whether it's a sensor or a battery, it's a resource." Look to the commercial world, he adds: "In the development part of the commercial world, when you're building the product, you have a strong incentive to be able to design-in interchangeability because you already know that the lifespan of any novel technology is perhaps a few years and certainly not decades. That drives you to develop technology in a certain way."

Progress towards the rapid production of platforms is already underway: Wardlaw's team has manufactured several large prototypes of 60 to 80-foot long, five-foot diameter platforms. Some are designed and 3-D printed in a couple of weeks and cost under a $100,000. "What's really cool is that, unlike the way things are traditionally done where I have to design my sensor or system to fit within an existing platform, introducing a multitude of platform constraints — everything has to be just right! — in this case, I can build the sensor and then build a platform around it!" says Wardlaw. "It provides the opportunity to get well inside of a build-test-learn-revise loop, similar to the observe-orient-decide-act loop. Which brings me back to deep learning: Now we're talking about the ability to actually put a ‘smart sensor' in the field, execute with it, find out what it learned, feed that information back, learn what needs to be revised, and try again. This will allow us to accelerate the learning process in a way that doesn't take 15 to 20 years."

The 3D printed Optionally Manned Technology Demonstrator, above, matches the design of a Mark 8 Mod 1 SEAL Delivery Vehicle (SDV), used to transport Navy SEALs and equipment under the sea. Credit: U.S. Navy photo by Chief Photographer's Mate Andrew McKaskle.

Wardlaw and his team are learning what autonomous vehicles can do, learning their capacity, and, given that capacity, learning how to make them more tactically sound and cost-effective for particular missions or "capabilities." "If we're smart and leverage AI systems and advance the manufacturing of platforms," says Wardlaw, "the side benefits include the ability to take advantage of less than optimal systems and integrate them. We can take power supplies that we're developing and put them in a different kind of propulsion unit, for example. And it's modular, so we can mix and match and try things out and do it quickly." For ONR, Wardlaw says, "That is huge."

Wardlaw's vision of capability-based sensing includes the dynamic reallocation of sensor resources to minimize excessive and distracting data: the point is to increase the sensors' relevant information content, or utility towards a given mission objective. This, says Wardlaw, is fundamentally an information theoretic problem: "I think we're just at the beginnings of how to capture and identify the different capacities that AI is going to bring; I think the more we play with it, the better we're going to get at it. We've gotten good at collecting data; we haven't gotten good at collecting information. The distinction being, that information is useful and relevant. What I'm suggesting is that instead of waiting until we've collected all this open-ended data and then trying to unwind and unfurl it on the back-end, once a sensor system learns the context from which to interpret data, it can begin to glean what's actually useful to collect. This gives us the opportunity to collect data in a much more relevant way."

Wardlaw - who does not remember a time that he did not see himself as an engineer or a scientist ("I was the kid in the neighborhood that people called on to fix their radios and TVs; for my musician friends, it was hand-built guitar amps, pre-amps and distortion units") - sees a critical need for photonics when it comes to security, particularly in the area of communications. Terrestrially, he says, communication has been relegated to the radio frequency part of the electromagnetic spectrum; within the oceans, acoustics is the norm. That, Wardlaw says, is about to change. "There's just so much data," he points out. "We need bigger pipes, and the telemetry advantage possible within the optics domain comes from its fundamental bandwidth."

A significant clue, comes from the fiber-optic industry: the same thing that drove the telecom industry into fiber-optics, is driving free-space communications into optics. Similarly, photonics is going to be used in increasingly disruptive ways, some of which Wardlaw is already involved in. "There are certain innovations that, when I share them with colleagues and say: ‘No, really, we're going to be doing that, like, next year,' they look at me like I'm nuts. For example, people have been trying to develop effective free-space power-beaming and power over fiber for a very long time; the idea has been around for many decades. And now that several completely separate programs have matured key supporting technologies, we can really do it! Laser technology is finally available to support such a crazy idea." In line with that, he says, is the solar industry with their high performance PV Cells. There's an inevitable progression of what Wardlaw describes as "a coincidence of supportive technologies" enabling this capability and many others to emerge: in this particular case, our photonics progress is expanding the power distribution infrastructure from copper cables to include power distribution via the magic of light. "Just imagine," says Wardlaw, "what future capabilities this might generate for us."