DARPA's biggest challenge-what's next for artificial intelli
DARPA's spectrum cooperation challenges may have proven that this idea is feasible, but artificial intelligence-managed spectrum sharing still has a long way to go.
"You have graduated from the spectrum hard knocks," said Paul Tilghman, project manager of the Defense Advanced Research Projects Agency (DARPA), in the final of the agency ’s Spectrum Collaboration Challenge (SC2) on October 23. The three-year competition has just ended. The top three teams are called on the stage, just like a song playing overhead, which sounds a bit like "Pomp and Circumstance".
It is not an exaggeration to describe this game as "hard-hitting". The 10 teams that reached the final, and other teams that have been eliminated in the previous rounds, have the task of proving something that has never been shown before. The challenge is to see if artificially managed radio systems can share the wireless spectrum more efficiently than static, pre-allocated frequency bands. They spent years playing in a DARPA RF simulator built specifically for the Coloseum.
At the end, the top team has demonstrated that their system can transmit more data on a lower spectrum than the existing LTE standard, and has demonstrated the amazing ability to reuse spectrum on multiple radios. In some finals, the radio systems of the five teams transmitted 200% or 300% more data than the current strict spectrum allocation. This is important because we are facing a vaguely visible crisis in the wireless spectrum.
But as Tilghman repeatedly emphasized during SC2, when the major challenge of DARPA is over, it does not mean that the technology is ready for market. These challenges are more to prove that a new technological concept is possible. In this regard, SC2 has succeeded in various standards.
As a result, great artificial intelligence-controlled radios can work together, sharing spectrum between them, and packing more data into a given amount of spectrum. But now that the DARPA challenge is over, where does technology go from here?
At two MWC Los Angeles teams, the Mobile Industry Conference where the finals were held, industry experts, employees of the Department of Defense (DoD) and Federal Communications Commission (FCC), and members of the top three teams participated in the discussions, and several appeared A clear theme.
An important point is that there is more room for cooperation. "Now the game is over," Tilghman said, "I know a lot of teams are looking forward to working with their competitors." This is what the team members shared in the group discussion. "We have very smart people here, and we haven't been able to ask‘ why did your radio do this? '”Said John Shea, the team member of the University of Florida GatorWings, who ranked first.
The competitive nature of this challenge also means that teams don't have to build truly collaborative and altruistic systems as they do in the real world. Ultimately, the top teams recognized that they must also get the most points to win, so the artificial intelligence system developed will also ensure that they maintain sufficient collaboration to meet the requirements of SC2. For example, GatorWing's system can identify when it can work with other best teams and enter a "competitive model." Similarly, the third-ranked Zylinium team's system also uses "Hulk Mode" and achieved the same effect.
But in addition to training artificial intelligence systems to be more collaborative, larger technical issues must also be addressed. Members of the GatorWings team discussed how they spent two years building and understanding custom software-defined radios for their systems. These team members all have 20 years of RF engineering experience. The winning team agreed that there is still a lot of work to do to improve the radio, including improving the system "brawn", that is, how well their signal can resist interference and get their own data.
The team also believes that training AIs in low-intensity simulations is valuable before putting AIs into a fully simulated environment like the Colosseum. Simulation will train AIs faster and reduce the time required to train AIs to be more accurate and robust.
Team members from the Department of Defense and the Federal Communications Commission discussed what to do to make players happy with the idea of sharing the spectrum. "We are being asked to share technology that is not normally shared," said Fred Moorefield, Deputy Deputy Chief Information Officer for Command, Control, Communications and Computer and Information Infrastructure Capabilities. He cited a broadband supply that must work with military satellite communications. The FCC's Office of Engineering and Technology Director Julius Knapp said that having competitors' service providers share the same spectrum in the real world can be tricky because collaborative radio work requires sharing of wireless communications that companies may be reluctant to disclose. information.
There is at least one chance that SC2's ending seems to be just in time: the deployment of 5G networks. Several frequency bands are currently being developed for the next generation of wireless technology, including the first millimeter wave band. But 5G also relies on the very valuable mid-range spectrum, which is still coveted and in short supply. Developing collaborative spectrum technology with 5G can make wireless power generation more successful.
In future wireless networks, we cannot predict when or how to achieve autonomous spectrum sharing. But research will definitely continue. Next, it is worth mentioning that the arena itself, the core test bed of SC2, will start a new life.
With the end of the MWC in Los Angeles, this test bench will be packed and trucked to Northeastern University in Boston, where it will become part of the National Science Foundation's Advanced Wireless Research Project platform.
keasysoft Online Lottery software development adheres to the service concept of making purchases more and more convenient and making services more and more perfect. After 8 years of continuous improvement of product functions and improvement of product advantages,keasysoft Lottery development we have been committed to the "industry digital experts" brand concept, focusing on Development and research of lottery systems. The unique advanced product development model ensures the success of each product of the customer from the three aspects of early demand collection, parallel development and after-sales service. Adhering to the business mission of "creating value for customers", based in Hangzhou, serving the world.