Robot Wars 362
EyesWideOpen writes "According to this New York Times article (free reg. req.) the Office of Naval Research is coordinating an effort to determine what it will take to build a system that will make it possible for autonomous vehicles (in the air and on the ground), or A. V.'s, to serve as soldiers on the battlefield. The project, called Multimedia Intelligent Network of Unattended Mobile Agents, or Minuteman, would consist of a network in which the highest-flying of the A. V.'s 'will communicate with headquarters, transmitting data and receiving commands. The commands will be passed along to a team of lower-flying A.V.'s that will relay them in turn to single drones serving as liaisons for squadrons of A.V.'s.' The article also mentions that the A. V.'s will have the ability to send high resolution color video as well as still photographs using MPEG-4 compression. Pretty interesting stuff."
What's the progress? (Score:4, Informative)
Maybe somebody from the project is reading this, and can provide some real information?
The M.I.N.U.M.A system is _NOT_ Autonomous (Score:2, Informative)
There is a top level project called "Intelligent Autonomous Agent Systems" of which this is part of. But there's nothing coming out of that which resembles T2 style aggresive AI controlled vehicles. Most of what they mean by autonomous, is the ability for the system to reconfigure itself if it loses an 'agent'. IE, and information node point. Another UAV could move from Group-A to Group-B to cover a lost eye-in-the-sky.
Although, I think there is room for truly autonmous agressive UAV. During desert storm, much of the day-day airborne offense took place in kill-boxes. They basically put a grid over the desert, and certain pilots or squadrons were told to destroy anything moving in grid X:Y. These boxes we're very much outside the 'Fire Support Coordination Line' meaning these air mission didn't need to be coordinated by someone on the ground. They were truly deep in enemy territory. When you run missions near troops the FSCL becomes the important factor. You can't target or shoot anything behind it (your computer won't let you either) Also, anything behind the FSCL requires a on-the-ground coordinator to give you the go ahead. I think we could see in 10 years roving aggresive UAVs that patrol grids and kill anything it finds in them. It's no different than what our pilots do now.
In fact, our humans pilots make mistake more than machines. There's famous video tape of an Apache captain taking out a Bradley and an M-113 at night, all capatured on his FLIR. He was providing FSCL support. His computer would not give him the green light to fire, he in fact had to override it in order to attack. His ground command did clear him for the shot verbally, telling them they had no vehicles in that area. There could be an argument that a mistake like that would not happen if it was a machine making the decision. I believe the real cause of that incident was the moving of the FSCL, and the airborne guys not getting the most recent FSCL coordinates (although his computer did have it).
-malakai