Back to news

August 12, 2020 | International, C4ISR

Lockheed Martin Ventures Scouts Next-Gen AI/ML Tech

"There's a massive scramble for autonomy engineers, software -- you name it," says Chris Moran, Lockheed Martin executive director.

By on August 11, 2020 at 4:16 PM

WASHINGTON: As defense primes scramble to meet DoD's insatiable demand for AI and machine learning (ML) tools, Lockheed Martin is investing in startups like Fiddler with next-generation tech to help operators understand how autonomous systems actually work (and don't work) in the field.

“I think what everyone is seeing is that, as you go toward deploying an AI/ML system, people start questioning well, how does this thing work?,” Chris Moran, executive director and general manager of Lockheed Martin Ventures explained in an interview yesterday. “And how do you know, how are you sure, it's making the right decisions? And how can you change those decisions if they're the wrong ones?”

“You want to make sure that these things are behaving,” he added. “Fiddler provides some of that insight into how artificial intelligence is working. They're in what's called ‘explainable AI' space, so they can reveal things about how the neural network was created, and how it's made decisions, right, which gives people a level of comfort,” Moran said.

Lockheed Martin Ventures, the mega-prime's venture capital arm, announced its investment yesterday in Fiddler, a two-year-old Palo Alto startup.

According to the joint press release, the two firms will work together “on the development, testing and scaling of Fiddler's technology in applying explainable AI in the defense and aerospace industries. ... At the heart of Fiddler's Platform lies AI Explainability, which provides continuous insights understandable by humans to help build responsible, transparent, and fair AI systems.”

AI/ML and autonomy are two of the key focus areas for Lockheed Martin Ventures, which buys equity shares in infant companies interested in selling to both the defense commercial marketplace, Moran said.

Not only is DoD racing to deploy AI/ML capabilities for everything from killer drone swarms to spare parts management, such systems are being integrated into almost every civil market sector from aerospace to agriculture — meaning an almost guaranteed return on investment, Moran explained. That return is then re-channeled into future investments.

“AI is such a hot topic right now that every company, not just the Lockheeds and the Boeings and the Northrops, but every single Fortune 500 company, maybe even every Fortune 1000 company, has realized ‘wow, I can simplify my tasks and move some of these mundane things into autonomous system, and thereby have people work on more complicated things that maybe are not suited for autonomy'. So, everybody is trying to do this. Everybody!” Moran enthused. “There's a massive scramble for autonomy engineers, software, you name it.”

Another next-generation autonomous technology development that has caught Moran's eye for possible future investment is the advent of what he called “autonomy factories;” that is, the ability to automate the process of building neural networks that can then build autonomous systems, autonomously.

“What's happened is that companies are starting to figure out how to automate autonomy — how do you autonomously create neural networks and machine learning systems?” he said with a chuckle of amazement. “You know, necessity is the mother of invention.”

Moran and his team of some six scouts have a $200 million fund to bet on newbie entrepreneurs and their technology. Currently, he said, Lockheed Martin Ventures has an investment in 40-odd companies, across 18 focus areas ranging from AI to rockets and propulsion systems to quantum science.

The focus areas are determined by a conclave, usually held in March, with Lockheed Martin's business units, as well as via the Ventures team's own knowledge of the startup ecosystem, he explained. In addition, Lockheed Martin Ventures haunts the increasing number of DoD, and especially Air Force, “pitch days” in hopes of finding matches for the aerospace prime contractor's interests.

As Breaking D readers know, “pitch days” are one of the new methods being championed by acquisition czar Will Roper as a way for the service to harness commercial innovation. And the Air Force is one of Lockheed Martin's biggest customers, if not the biggest if you count space acquisitions.

The Ventures team invested in 10 startups last yea,r including one discovered at an Air Force pitch day, and is on a path to adding another 10 to its portfolio this year, Moran said. Finding those winners is an intensive process that involves scouting 700 to 1,000 startups per year, he explained.

Once a startup is chosen, Lockheed Martin Ventures gives it an opportunity to pitch ideas/products/services across all interested Lockheed Martin business areas.

“We have an internal, if you will, we call it ‘demo day,' that we're holding this week, and right now I think there are well over 100 Lockheed Martin engineers and technologists set up to listen to 12 or 13 of the portfolio companies that we've invested in the last year,” he said. “That's part of what we do as a service inside the company. And hopefully out of those discussions and presentations, there are further collaborations.”

Moran explained that those collaborations can include software licenses or contracts for services — and even, once a startup has established a larger market presence, traditional subcontractor ties.

And while investment allows Lockheed Martin Ventures to get in early on the startup's expertise and tech concepts, the prime contractor is not seeking to tie the hands of the entrepreneurs regarding clientele. Instead, he said, the objective is to grow the startup into the overall defense and aerospace industrial base.

“We're creating, eventually, a market for them,” Moran said. “And it's kind of a weird dynamic, but we give them money, and then my group goes off and works for them inside the company — so we're paying them for us to work. It's weird, but ... that money goes to supporting the small companies grow and scale so that they'll be around for when a Lockheed Martin or any large company wants to use their tools and services. So we look at it as a win-win.”

https://breakingdefense.com/2020/08/lockheed-martin-ventures-scouts-next-gen-ai-ml-tech

On the same subject

  • Navy Exercises $84.7 Million Option With Boeing For Three MQ-25A Stingrays

    April 3, 2020 | International, Naval

    Navy Exercises $84.7 Million Option With Boeing For Three MQ-25A Stingrays

    By: Ben Werner The Navy exercised contract options with Boeing worth $84.7 million to buy three MQ-25A Stingray unmanned aerial refueling tankers as part of a modification to a previously awarded contract. The three MQ-25s covered by the contract options are to be completed by August 2024, according to the contract announcement released by the Pentagon late Thursday. For more than a decade, the Navy has conducted an on-again, off-again development process to build what will be the Navy's first unmanned carrier-based tankers. In 2018, Boeing beat Lockheed Martin and General Atomics to land the $805 million contract to build the first four MQ-25As. The Navy anticipates integrating MQ-25 aircraft into the carrier air wing for initial operational capability by 2024, according to a Navy's fact sheet about the program. The Navy's desire for an unmanned aerial refueling tanker is born out of necessity. Currently, between 20 and 30 percent of a carrier-based air wing's Super Hornet fleet is dedicated to aerial refueling operations. Introducing the MQ-25A Stingray will free those Super Hornets for strike missions. The MQ-25A is expected to deliver up to 15,000 pounds of fuel at 500 nautical miles. In September, Boeing conducted the first flight of its unmanned aerial refueling test aircraft the T1. Boeing tested both the T1's handling and the Navy's ability to fast-track the long-stalled program. Boeing first unveiled photos of the T1 prototype months before winning the contract award. Having the T1 available enabled the Navy and Boeing to perform tests much earlier in the development process than would be typical for a new build. The type of testing performed in September, a year after the contract award, usually doesn't start until the first engineering development models (EDM) are built. With the MQ-25 program, the first EDM airframes are expected to be ready for testing by the end of 2021, according to the Navy. Boeing previously announced plans to hoist the T1 onto an aircraft carrier and conduct deck handling tests this year. Boeing plans to shift to using the first EDM airframes for testing in 2021. The first carrier-based tests and sea trials are expected to occur in 2022 and 2023, the Navy previously stated. https://news.usni.org/2020/04/02/navy-exercises-84-7-million-option-with-boeing-for-three-mq-25a-stingrays

  • U.S. Navy Commissions Littoral Combat Ship 15 (Billings)

    August 5, 2019 | International, Naval

    U.S. Navy Commissions Littoral Combat Ship 15 (Billings)

    KEY WEST, Fla., Aug. 3, 2019 /PRNewswire/ -- The U.S. Navy commissioned USS Billings (LCS 15) – the nation's eighth Freedom-variant Littoral Combat Ship (LCS) – in Key West, Florida. This milestone places the ship, built by the Lockheed Martin (NYSE: LMT) -led team into active service. "Billings was designed to operate and adapt to a rapidly changing environment," said Joe DePietro, vice president of Small Combatants and Ship Systems at Lockheed Martin. "She is equipped and ready for today's threats and easily modifiable to meet the threats we may not even be aware of yet. Our team is confident Billings will be what the Navy needs when the fleet needs it." Unique among combat ships, LCS is designed to complete close-to-shore missions and is a growing and relevant part of the Navy's fleet. It is fast — capable of speeds in excess of 40 knots. It is automated — with the most efficient staffing of any combat ship. It is lethal — standard equipped with Rolling Airframe Missiles (RAM) and a Mark 110 gun, capable of firing 220 rounds per minute. It is flexible — with 40 percent of the hull easily reconfigurable, integrating capabilities like the Longbow Hellfire Missiles, 30mm guns, and manned and unmanned vehicles targeted to meet today's and tomorrow's missions. "Having now commanded two freedom class LCS variants, I would like to report that these ships are truly impressive and will fit well in the niche they have been designed for," said LCS 15's Commanding Officer, Commander Nathan Rowan. "They are fast, maneuverable, and their weapon systems are some of the most accurate I've witnessed on any platform of which I've previously served." There are seven ships in various stages of production and test at Fincantieri Marinette Marine in Wisconsin, where the Freedom-variant LCS is built. The next Freedom-variant in the class is LCS 17, the future USS Indianapolis, which was delivered in late July. "On behalf of the proud 2,000 men and women who transform flat steel into a fast, agile surface combatant," said Jan Allman, CEO of Fincantieri Marinette Marine, "we are honored to support the U.S. Navy, and we congratulate the outstanding crew of the USS Billings." Multimedia assets are available here: Social media video: https://vimeo.com/351803068 B-roll: https://vimeo.com/351805414 Photos: https://www.flickr.com/photos/143371902@N04/albums/72157710070352016 For more information, visit www.lockheedmartin.com/lcs. https://news.lockheedmartin.com/2019-08-03-U-S-Navy-Commissions-Littoral-Combat-Ship-15-Billings

  • Laying the groundwork: US Army unveils rough plan to formalize robotic combat vehicles effort

    September 22, 2020 | International, Land, C4ISR

    Laying the groundwork: US Army unveils rough plan to formalize robotic combat vehicles effort

    Jen Judson WASHINGTON — The U.S. Army is preparing to enter into programs of record for light- and medium-class robotic combat vehicles in a few years, with plans to award separate contracts for a lead systems integrator for each program, according to the service's robotic combat vehicles product manager. The service wants to field a light, medium and heavy robotic combat vehicle, and it is experimenting with technology and how the vehicles might fit into future formations through the Army's Next-Generation Combat Vehicle modernization office. Developing NGCV capability is the second-highest priority for the Army. The plan is to make a decision to move the Army's RCV-Light out of technical maturation and into the engineering and manufacturing development phase in the second quarter of fiscal 2023. The service would do the same with the RCV-Medium program in FY24, Lt. Col. Chris Orlowski said Sept. 10 during a conference hosted by the Association for Unmanned Vehicle Systems International. Orlowski noted that the potential RCV-Heavy program of record would fall behind the start of the medium and light programs by “a pretty significant margin.” At the same conference during a separate keynote presentation, Bruce Jette, the Army's acquisition chief, said decisions were made that “put pieces in place that will establish a formal program for robotic systems with the [program manager], not just following the tech base but in fact being the centerpiece for building light and medium systems inside of an architecture that fits within the entire operational vehicle architecture. In fact, it will leap over into the aircraft as well. It will be everything from driving to operations of the vehicle to visual sensing to probably a broader array of technical capabilities that you may not have even thought possible.” The Army plans to award a contract to a lead systems integrator that will combine the RCV's control station, network, platform, software and payloads, Orlowski said. Anticipated government-furnished equipment for the programs would include autonomy software, radios, war-fighter machine interface software, aided target recognition software and lethality payloads “minus the turret,” according to Orlowski. Other potential government-furnished equipment could be a tethered drone; assured position, navigation and timing technology; hostile fire detection; and other vehicle protection systems, he said. As the Army ventures into developing robotic vehicles that don't just do the dull, dirty and dangerous work, “the biggest thing is going to be software development, improving autonomous and automation software,” he added. “Teleoperation is nice; it works OK if you've got the right radios and the right environments, but long term, when those environments become tested, I think teleoperation will be less viable and we will have to really push the automation and autonomy on these platforms. But also, that being said, there is always going to have to be some soldier interaction with the platforms. How do we improve that interaction for the soldiers, reducing that cognitive burden?" he said. “I know everybody likes cool, big, awesome robots, but it's really a software thing that is going to make these things go, so anything that is kind of tied to software I think is a critical technology in my view,” he added. The Army's rough acquisition strategy for the robots calls for a first unit to receive RCV-Light vehicles in FY28 and a first unit to receive RCV-Medium in FY30, according to a slide from Orlowski's conference presentation. The Army recently wrapped up its first phase of experimentation with RCV-Heavy surrogates fashioned out of M113 armored personnel carriers at Camp Red Devil on Fort Carson, Colorado, which added complexity to an ongoing evaluation of the government-developed platforms. The Army also awarded contracts to a Textron and Howe & Howe team to build an RCV-Medium prototype, and to a QinetiQ North America and Pratt & Miller team to build the lighter version late last year and early this year. Those are being built now. Orlowski stressed those prototypes are being built “primarily to support future planned experimentation” in FY22 and FY24 and “in support of defining and informing requirements for the RCV program of record.” He added there is no plan to transition any of those systems into any type of limited fielding. “They are not designed for that. They are designed for a campaign of learning,” he said. Now that the first major experiment is done, the Army plans to build up to a company-level operation in the first quarter of FY22 at Fort Hood, Texas, with four RCV-Medium and four RCV-Light prototypes. While the experimentation at Fort Carson with RCV-Heavy was focused on cavalry operations where the robots served more in a scout mission and proved they could be effective in a reconnaissance and security role, the experiment in FY22 will move the robots into more of an attack-and-defend role. While the Army has to officially complete a critical technology assessment from the Fort Carson experimentation, Orlowski said the autonomy software “needs to improve.” The experimentation in FY22 will focus on improvements and the Army will work with industry partners to “improve that tether,” he said. “It needs to be robust in contested environments, which we haven't fully explored yet.” The service will also need to look at some alternate communications pathways between the control vehicle and the robots. Aided target recognition needs more maturity, Orlowski explained, “especially on the move to support the platforms.” Anything that reduces the soldier interaction with the platforms will also be incorporated, “and there are other things that soldiers asked for, which when we are ready to release we will. [The ideas from the feedback] were pretty perceptive," he said. "So how critical that becomes, we will see.” https://www.defensenews.com/land/2020/09/21/laying-the-groundwork-us-army-unveils-rough-plan-to-formalize-robotic-combat-vehicles-effort/

All news