MSNBC reports that cars are becoming increasingly autonomous.
The current set of semi-autonomous safety features can quickly combine into something more. For example, a car could use Lane Keep Assist and Adaptive Cruise Control together to drive itself under highway conditions, sticking to one lane and not hitting the car in front. The next step is to expand these capabilities. Adaptive Cruise Control currently works only over 25 mph, but the next version (called Full Speed Range ACC) lowers that number to zero so that cars can begin to handle traffic jams in the city.
The problem with this is that each car is autonomous. TAE holds that you cannot build a network of completely autonomous robots without an immediate cascade failure. In mathematics we call this "one equation with two unknowns." You need a second equation to drive a value for each unknown. Or if there are three unknowns, you need three equations. Or if there are five hundred thousand cars commuting one morning in Nashville, you need fifty thousand equations...in each car.
Or, conversely, you could have a single control mechanism running all the cars. Suddenly you go from five hundred thousand organisms to a single organism with five hundred thousand legs. A superorganism. And these have already been on Earth for millions of years.
Take this example. If two fully robotic, fully autonomous cars are going down the highway, and up ahead there is a traffic accident, they will just slow and crawl through traffic like the rest of the cars around them. However, if a central architect is running the flow of traffic, and an accident occurs, it can automatically divert one of the two cars onto other highways or side roads, and the traffic is lessened so that both cars potentially do not lose time on their trips.
A reader might argue that the cars could still be autonomous, while having the advantage of the traffic diversion techniques. That is true, and is probably a logical step that will occur in the gradual evolution of autonomous vehicles. An accident occurs, the vehicle informs you of it and you elect to take a different route. However, you are already giving up your autonomy.
And consider this, would a traffic diversion protocol be better served if people had the option to ignore it's suggestions? Or should an autonomous vehicle, for your expedience (and others), just divert you around traffic accidents without your input?
A better system would be to mimic, on a much more intricate and low-altitude scale, the system used by Air Traffic Control to keep airplanes from colliding in mid-air. Micro-management of airplanes is done by a pilot, he/she takes off, climbs to altitude, changes direction and pitch, yaw, etc., descends, and eventually lands. However, the actual flightpath of the plane is determined by an Air Traffic Controller sitting in a dark room at several locations across the nation, who is specialized in calculating vectors and intersection points, and carefully keeps planes from ever coming within a few miles of one another.
Why is this a good idea? Other than the obvious reduction in airplane collisions, we are returned to the "two floating points" problem. In lieu of ATC vector-guidance, a pilot, up in the sky, might have a radar screen and see a nearby airplane approching. So the pilot might adjust their course. Which might compel a third airplane pilot to adjust his/her course. Which might compel a fourth pilot to adjust, and so on and so forth until we come around to the original pilot, who has to move to avoid the nth airplane, and the circle continues. If this free-floating directional guidance were to happen, planes would zigzag across the sky and waste valuable time and fuel. They'd hit weather patterns unexpectedly, be unable to avoid turbulence, and potentially a large number of airplanes might reach their destination at the same time, snagging an airport in a quagmire of landing airplanes.
This system would work for cars. The autonomous driving systems, like those suggested in the MSNBC article, could provide micromanagement of the car in its current location. But an automobile traffic control system, sending simple navigational data to the cars to optimize the entire traffic system would produce the most efficient system for traffic. The traffic system could say "take highway 50 to Smart Street and take Smart Street to 56th Avenue." and the autonomous vehicle could then do so. Should an accident occur on Smart Street, the traffic system could update the car: "take Alternatate Avenue 2 miles then turn south and divert back to Smart Street, continue on original path from there."
Basically, what I am suggesting, is that everyone on earth has a GPS device in their car and all the GPS devices are linked to a central optimization database, which runs on 3 rules:
TAE's Three Rules of Traffic Robotics. 1. Provide the shortest (time) path possible from point A to point B for each car.
2. Minimize the number of cars at any given point at any time, without violating rule 1.
3. Do not divert a car's path mid-travel unless the new path is shorter (time) than the existing path.
You can read my other thoughts on cars that drive themselves here, here and here. As you can imagine, I feel strongly about this.
_
0 comments:
Post a Comment