How a defense expo reflects troubles for South African military

Sur le même sujet

• 

  19 septembre 2018 | International, C4ISR

  Differentiating a port from a shipyard is a new kind of problem for AI

  By: Daniel Cebul It's well known that satellites and other intelligence, surveillance and reconnaissance platforms collect more data than is possible for humans to analyze. To tackle this problem, the Intelligence Advanced Research Projects Activity, or IARPA, conducted the Functional Map of the World (fMoW) TopCoder challenge from July 2017 through February 2018, inviting researchers in industry and academia to develop deep learning algorithms capable of scanning and identifying different classes of objects in satellite imagery. IARPA curated a dataset of 1 million annotated, high-resolution satellite images aggregated using automated algorithms and crowd sourced images for competitors to train their algorithms to classify objects into 63 classes, such as airports, schools, oil wells, shipyards, or ports. Researchers powered their deep learning algorithms by combining large neural networks, known as convolutional neural networks (CNNs), and computers with large amounts of processing power. The result was a network that, when fed massive amounts of training data, can learn to identify and classify various objects from satellite imagery. By combining a number of these networks into what is called an ensemble, the algorithm can judge the results from each CNN to produce a final, improved result that is more robust than any single CNN. This is how a team from Lockheed Martin, led by Mark Pritt, designed their deep learning algorithm for the challenge. Pritt explained to C4ISRNET, that he and his team developed their CNN using machine learning software and framework from online open source software libraries, such as Tensor Flow. Earning a top five finish, the algorithm designed by Pritt's team achieved a total accuracy of 83 percent, and was able to classify 100 objects per second. Pritt said that with fully functioning algorithm, this software could take an image recognition task that takes a human an hour to complete and reduce the process to a few seconds. The team's algorithm excelled at identifying classes with distinctive features, and successfully matched nuclear power plants, tunnel openings, runways, tool booths, and wind farms with accuracies greater than 95 percent, but struggled with more indiscreet classes such as shipyards and ports, hospitals, office buildings, and police stations. “Usually when you develop an algorithm its nice to see where it succeeds, but you actually learn the most where you look at where the algorithm fails or it doesn't do well,” Pritt said. In trying to decipher why the algorithms struggled, Pritt said the competitors suggested that some objects simply don't have any distinguishing features from the point of view of a satellite image for the algorithms to recognize. “Maybe the most important ingredient you need for these new types of algorithm to work is the dataset because these algorithms require a great amount of data to train on,” Pritt explained. “It's kind of analogous to the way a human will learn in childhood how to recognize things. You need lots of examples of what those things are and then you can start to generalize and make your own judgments,” he said. But even with large amounts of training data that is correctly labeled, it is also possible the deep learning technology of today cannot reach the higher levels of intelligence to recognize nuanced differences. For example, Lockheed Martin's algorithm confused shipyards and ports 56 percent of the time. Pritt said that people “look at an image and they can tell that it's a port or a shipyard, they are usually looking at very subtle things such as if there is a ship in dry dock or if there is a certain type of crane present. They are looking for details in the image that are maybe higher level or more complicated than what these deep learning algorithms can do right now.” However, the fact that these algorithms cannot do everything should not dismiss the significant contribution they could provide to the defense and intelligence community. Hakjae Kim, IARPA's program manager for the fMoW challenge, said the benefits of this technology could extend far beyond faster image processing. “I want to look at it more in the perspective that we can do things we weren't able to do before,” Kim said. “Because its technology that we are now able to do x, y and z, there are more applications you can create because with the human power it is just impossible to do before.” Kim and Pritt stressed managing expectations for CNN-based artificial intelligence. “This is a real technology that will work, but it also has limitations. I don't want to express this technology as a magic box that will just solve everything magically,” Kim said. “I don't want the users in the field to get disappointed by the initial delivery of this technology and say 'Oh, this is another technology that was oversold and this is not something we can use," he added. Part of managing our expectations for AI requires recognizing that although intelligence is in the name, this technology does not think and reason like humans. “A lot of the time we think that because we use the term AI, we tend to think these algorithms are like us, they are intelligent like us,” Pritt said. “And in someways they seem to mimic our intelligence, but when they fail we realize ‘Oh, this algorithm doesn't really know anything, [it] doesn't have any common sense.'” So how are IARPA and Lockheed Martin working to improve their algorithms? For IARPA, Kim's team is working on updating and maintaining their dataset to ensure algorithms have the most up to date information to train on, ultimately making the CNN-based algorithms easier to trust. “[S]ubtle changes in the area mess up the brains of the system and that system will give you a totally wrong answer,” Kim explained. “So we have planned to continuously look over the area and make sure the algorithm we are developing and reassessing for the government to test on and use to be robust enough for their application," he furthered. Work is also underway at American universities. Kim described how a team of researchers at Boston University are using the fMoW dataset and tested algorithms to create heat maps that visualize what part of the image algorithms are using to classify objects. They've found that sometimes it is not the object itself, but clues surrounding the object that aid most in classification. For example a “windmill that actually shows a shadow gives a really good indicator of what that object is,” Kim said. “Shadows show a better view of the object. A shadow is casting the side view of the object over on the ground, so [BU's heat map algorithm] actually points out the shadow is really important and the key feature to make the object identified as a windmill.” But don't expect these algorithms to take away the jobs of analysts any time soon. “I think you still need a human doing the important judgments and kind of higher level thinking,” Pritt said. “I don't think AI will take away our jobs and replace humans, but I think what we have to do is figure out how to use them as a tool and how to use them efficiently, and that of course requires understanding what they do well and what they do poorly," he concluded. https://www.c4isrnet.com/intel-geoint/2018/09/18/differentiating-a-port-from-a-shipyard-is-a-new-kind-of-problem-for-ai

• 

  16 décembre 2021 | International, Aérospatial, Naval, Terrestre, C4ISR, Sécurité

  How Will The Next Pentagon Strategy Deal With China?

  U.S. Defense Department strikes a balance between raising tensions with Beijing and its case for urgent modernization.

• 

  1 octobre 2019 | International, Aérospatial

  Is rapid prototyping the key to space?

  By: Nathan Strout Space technology is developing so fast that by the time the tech makes it through the acquisitions process and into orbit, it's practically obsolete. “The technology is moving at such an accelerated pace and these technologies are on such a steep trajectory that the traditional acquisition system just frankly can't keep up,” explained Ken Peterman, president of government systems for Viasat. To combat this trend, the U.S. Air Force's space acquisition arm, the Space and Missile Systems Center, is focused on quickly building prototypes as a way to speed up development and bring nontraditional companies into the Pentagon's space. Using tools such as other transaction authorities, Section 804 — a rapid acquisition approach that aims to field capabilities within two to five years — the Space Enterprise Consortium and Air Force pitch days, SMC is encouraging an acquisitions model focused on prototyping over the slower, more cumbersome Pentagon procurement procedure. The missile warning example Under the Pentagon's normal acquisition process, it could take the better part of a decade — or longer — from contract award to launching a satellite into orbit. Consider the Next Generation Overhead Persistent Infrared satellite system. OPIR is supposed to replace the Space-Based Infrared System as the nation's premier early warning missile detection and tracking satellite system, providing significantly more capacity than the current constellation. But under traditional acquisitions, OPIR wouldn't be available until nine years after contract award. That wasn't going to cut it. So service leaders took a new approach. Instead of going through the traditional acquisition process, the Air Force would adopt a rapid prototyping approach to accelerate development. And rather than develop large, exquisite satellites meant to last generations, new satellites would be smaller, less expensive and built for replacement in three- to four-year cycles. “We have existing programs of record ... that continue to take the four or five or six years or whatever that we originally contracted for. I think our goal going into the future is to get more on a three- to four-year cycle for our satellites, not just in production but also in terms of their amount of principal time on work,” said SMC Commander Lt. Gen. John Thompson. Lockheed Martin used this approach to build three OPIR satellites. Instead of subcontracting with one company to build the satellites' payloads, Lockheed Martin is having two teams compete to build the best OPIR payload. This not only mitigates the risk of failure by having two competing prototypes, but also moves to a fail-early model where a successful prototype is less likely to lead to issues later on. Applied to OPIR, rapid prototyping has cut the projected timeline for the project in half. “It is harder to move fast if you haven't done the underlying innovation, prototyping and technology development that allows those systems to go forward. So even when you look at things like Next Gen OPIR, which is not a three- to four-year cycle, we're bringing that from what would have been 108 months down into the five-year timeline; that was the fuel,” said Col. Dennis Bythewood, program executive officer for space development at SMC. Hither the Space Enterprise Consortium OPIR is one example of the rapid prototyping approach the Air Force is taking with space procurement. But perhaps the best example of SMC's rapid prototyping approach is the Space Enterprise Consortium, or SpEC. Established in 2017 through an other transaction authority, or OTA, SpEC is an organization comprised of 325 members who can apply to build space-related prototypes. To date, the Air Force has issued more than 50 awards to the group to develop prototypes, ranging from a ground system for OPIR to a new space vehicle that could extend the Link 16 network to beyond line-of-sight communications. Often, the consortium awards multiple contracts for one project, allowing companies to compete to produce a viable technology. Although not every prototype succeeds, the approach allows for failure earlier in the development process than the traditional acquisition model. “[This strategy] allows individual components to fail, but us to continue to move forward. And that's a place where you can see competitive prototyping with multiple vendors going head-to-head. I don't know which one will be successful when we start, but there's a much higher likelihood that I'm going to end up with multiple success at the tail end,” Bythewood said. Because the consortium specializes in OTAs, it's able to open the door to nontraditional companies that don't have the time or resources to go through the regular acquisition process under the Department of Defense. “OTAs are good in the sense that they're a lot more flexible, they're not hard contracts — they're just agreements. And in that respect, they attract nontraditional suppliers, people who haven't been working with DoD and don't want to recall all their accounting systems, so that they can comply with all the cost-reporting requirements for DoD and everything like that. So for a lot of companies, they're attractive,” said Cristina Chaplain, director of the Government Accountability Office's contracting and national security acquisitions team. Small startups and venture capital-funded companies are a big part of commercial growth in space. But for a long time, these companies have been frustrated in their attempts to engage with the Pentagon on space, Chaplain said. OTAs, and the consortium in particular, knock down barriers between the DoD and small, commercial, space-focused companies. Of course, working with less traditional companies also opens the door to increased risk. These companies aren't necessarily familiar with the way the Pentagon does business. And even if OTAs offer flexibility, working with the government be a major challenge. That's why SpEC was designed with mentorship in mind. In the SpEC framework, there's room for these small companies to partner with larger, more established players. Kay Sears, Lockheed Martin's vice president and general manager for military space, explained that defense contractors can work with SpEC to find innovative startup companies that need help bringing their new technologies to bare for the Pentagon. The result is a symbiotic relationship, where prime contractors such as Lockheed can take smaller companies under their wing as they navigate the complex world of the DoD, while the startups can help Lockheed innovate. “We're not asking those companies to become defense companies, we're asking them to actually stay commercial and stay motivated to their original business plan, but to work with us and we can mentor them to help develop that technology,” Sears said. “So we have to find those nuggets of commercial capability and commercial innovation, and then bridge that into the mission understanding that we have and the mission systems that we can contract (for) and deliver.” ‘The darker side' While the Air Force is quick to tout the expected benefits of the SpEC approach, there are potential downsides. For one, transparency. “The darker side is that it's harder to have good management and oversight if you're not requiring all the same things from the contractors. You're not getting the same kind of reporting,” Chaplain said. The Government Accountability Office can help hold the Pentagon and contractors accountable over the long lifetime of a program contract, tracking spending increases, delays and failures. And the GAO is able to provide some oversight for OTAs, however it's more difficult than programs going through the regular acquisitions process, Chaplain explained. Another problem is funding. The rapid prototyping approach requires more money up front and a less risk-averse approach. Next Gen OPIR will be a test run for whether Congress can get on board with that approach for space. As the Air Force sped up OPIR's timeline with rapid prototyping, it created a significant increase in their near-term budget. For fiscal 2020, the Pentagon asked for $1.4 billion for the program. That's a $459 million increase over what was projected for FY20 during the last budget cycle. The House has balked at that amount, authorizing $1 billion of the requested funding. The Senate Appropriations Committee has taken the opposite approach. Not only did Senate appropriators vote to fully fund the request; they threw in an additional $536 million to fully fund the program. As the senators noted in their report on the bill, OPIR will serve as a test case for whether Congress will support SMC's rapid prototyping approach. “The Committee believes the program will be a[n] exemplar for rapid acquisition of space programs, whether the program succeeds or fails,” the report read. “Failure will have implications for Congress's willingness to fund future programs using the National Defense Authorization Act section 804 rapid prototyping and fielding authorities for similarly large, or even middle tier programs, for years to come. Alternatively, if the program is to have any chance of success, the [Defense] Department cannot continue to rely on reprogramming requests for its funding.” The once and future SpEC Even as the fight over OPIR funding continues in Congress, the Space Enterprise Consortium and its funding has grown by leaps and bounds. “It has been a vast success story for [SpEC]. We began that contract with a $100 million ceiling, which meant that we could execute many different actions within it up to about $100,000. We took that five times higher within the first year,” Bythewood said. And SMC seems keen to build on that approach. On Aug. 20, the Space and Missiles Systems Center issued a request for information expressing an interest in re-competing the SpEC OTA agreement. This new SpEC would have a $12 billion ceiling over 10 years. “We're not going to be awarding a $1 billion contract within SpEC OTA. We're looking at having smaller competitive prototyping efforts that get our products off on the right start in order to deliver capabilities sooner. So if there's a fear ... that we're going to be executing huge programs of record under the SpEC OTA vehicle, that's a kind of [unfounded] fear,” Thompson said. Another example of the rapid prototyping initiative is the Air Force's new pitch days concept, where on designated days, companies can present new technologies to the government and potentially win a Section 804 contract within minutes. The Air Force has been holding pitch days this year for a variety of platforms. The Air Force will be holding its first “Space Pitch Day” from Nov. 4-8 with a focus on launch systems, data mining, space visualization and space communications. Whether it's SpEC, pitch days or working closely with contractors, it's clear SMC sees rapid prototyping as the way forward for military space. “We recognize that when you try new things, some will work great, some will work moderately well and some you might fail fast on. But that's OK because clearly we need to do things differently,” said Peterman of Viasat. “We applaud the kinds of things these senior leaders are doing to try to drive change, get these cutting-edge capabilities into the war fighters' hands as quickly as possible.” https://www.c4isrnet.com/battlefield-tech/space/2019/09/30/is-rapid-prototyping-the-key-to-space/