10 October 2005

Robot Drivers

A robot controlled Volkswagen SUV has successfully traversed a 132 mile course set out by the Defense Department in the Mojave Desert in a little under seven hours. The implications go far beyond the two million dollar prize won by the Stanford University team that designed it for about $500,000.

The field of robotics and artificial intelligence has been around for a long time, but it is just starting to come into its own. While an unmanned combat aerial vehicle was used a few times between the Gulf War and the Iraq War for combat missions, and they have been used to a limited degree in the Iraq War, as have a few early versions of remote controlled ground combat platforms such as the Talon robot, the real impact that robots will have on warfare has not yet been realized. But, this will happen.

In the Air Force, it is already well known that drones can make very good reconnaissance and bombing platforms. The reconnaissance function is the primary role that drones have had in the Iraq War, and it has become standard operating procedure for units as small a platoons (fewer than fifty soldiers) to have their own eyes in the sky to ferret out ambushes and get the lay of the land ahead of them. Limited use has been made of drones for close air support, with a missile armed drone firing into a building from which enemy gunfire is coming, or enemy vehicles.

The ability of drones, unburdened with life support systems required by human pilots, to linger over an area where troops need air support, dropping bombs on demand at targets identified with global positioning satellite coordinates or laser pointers provided by forward observers on the ground, is obvious. Likewise, drones are perfect for flying over pre-selected targets, dropping bombs on them and returning to base. The tasks are fundamentally ministerial, requiring no discretion on the part of the pilot. Drones have already done this in tests, and indeed, have gone one step further, deviating from their pre-planned mission to autonomously avoid anti-aircraft fire, to locate its source, and to, with permission from the command post, fire back and destroy that source before going on with its mission.

In order to assuage concern from the “fighter mafia” that runs the Air Force, drones have not yet officially been built or even designed for air to air combat, but few people doubt that drones would have an edge here as well. Why? They are smaller than comparable manned aircraft and are also capable of performing maneuvers that would kill or render unconscious a human pilot with high G forces. If flight simulators used to train pilots can run cyber planes in air to air combat, there is little doubt that software could also be designed to implement similar tactics in real life.

In the Navy tests are already underway to develop unmanned surface ships and submarines. Singapore has already adopted a U.S. beta version for its own Navy to patrol its coastal areas. Unmanned boats seem ideally suited for the routine vigilance required of boats searching for mines or enemy submarines in mostly empty waters. Even on manned ships, automation is making it possible to run ships that conventionally would have required a crew of three or four hundred sailors with fifty. Critics argue that Naval ships need large crews to provide damage control when a ship is hit. But, it is hard to justify paying salaries for 250 extra sailors and putting those sailors in harms way, simply to serve as a fire department for the ship on the rare occasions when a Naval ship is hit, but can be saved with damage control efforts of a type that can only be provided by a large crew of sailors. Moreover, even if that is possible, the crew mix on such ships would likely shift from 300 sailors to 50 sailors and 250 Marines, who could be used for a variety of roles far beyond damage control. Were the entire Navy to reduce its personnel needs by 5/6ths, the Navy could shed more than 250,000 sailors without any loss in effectiveness.

Of course, the latest developments in the Mojave are most pertinent to the Army and the Marines. One can easily imagine fully automated tanks replacing the existing models with crews of three or four men, with half a dozen or so of them led by a single commander. If you can have a ground vehicle that can drive itself autonomously, it doesn’t take much more software to fully automate self-propelled artillery batteries. One can imagine supply convoys in which only one or two of the vehicles have soldiers operating them, greatly reducing the risk of harm from roadside bombs. One can imagine field ambulances that autonomously make their way across a battlefield to retrieve injured soldiers, without putting more soldiers at risk in the process. One doesn’t even have to imagine R2D2 sized robots, remotely operated from a hundred yards away, that can be thrown in the window of a building where armed opponents are believed to be located searching the building for armed persons and killing them, so that soldiers don’t have to be put at risk in this high risk type of mission performed thousands of times a week in Iraq.

The civilian implications may be equally great, however. Imagine tractors and combines that leave the barn on schedule, deliver themselves to the appointed field for the day, and carry out their tasks on a GPS guided course. This isn’t so far from reality. Modern tractors are now human manned mostly to get them to the proper field and to turn around at the end of the field to start the next row, many already drive on GPS guided autopilots from one end of the field to another. I’ve heard of a farmer, possibly apocryphal, who went back to the tractor dealer demanding additional reading material as he had so little to do in the field that he had read the owner’s manual cover to cover three times already. Robot tractors and combines could turn sixteen hour days for two week’s straight, a few times a year, into a one time effort to program the routes, accompanied by a day or two, a few times a year to adjust the pre-existing settings and fill the machines will oil, gas, seed, and agricultural chemicals, all without leaving the home place. Alternately, a single family farm might be able to manage 50,000+ acres of crops with a dozen or two robotic farm vehicles at their disposal. This would put 90% of farmers out of business in the United States, but would also restore farms to profitability in the international marketplace.

Far simpler software could guide long haul trucks up and down the nation’s interstates, particularly with a modest investment in radio frequency ID (RFID) guideposts to line the highways. Add crash collision radar and specialized turnoff exits on the fringes of every major city and robots could haul most of the intercity cargo in America.

1 comment:

Anonymous said...