The Self in Self-Driving Cars
The Rand Corporation had a post referenced in the science news aggregator RealClearScience about self-driving cars which is a subject that greatly interest me. Rand questioned what is the current state of the whole situation, and I have to agree with them – as far as they go. The major point of the Rand article was that even if self-driving cars where tested for way more miles than they have currently run, it still would not be enough to determine their safety on the road compared with human driven cars.
One major subject not being mentioned about self-driving transportation is that it has already existed for hundreds of years. It was called the horse. All kinds of stories exist of drunks or just overly drowsy riders going out and getting in a carriage or “back in the saddle” and being safely transported home. And why? Because the horse knew the way home and was neither drunk or asleep.
Also – and this is the key to the self-driving car discussion – the horse did not use the riders form of intelligence to get home safely. It used its own (horse sense?) – so too should self-driving cars. And while we are on the subject of going from horse to automobile form of transportation, infrastructure had to change when we went from one form to the other. This has been true all through history: gas lighting to electricity, sailing ships to steam, letter writing communication to telegraph to telephone to radio and TV, and let’s not forget the Internet. So too does this need to happen when going to self-driving cars. I would think a group like the Rand Corporation should lead the way instead of describing what is wrong.
Self-driving cars should not try to observe the road the same way a human does. Trying to scan road markings, the general direction of the blacktop, gutters, curbs, or even road signs meant for humans is a failure waiting to happen, as the Rand posting so aptly describes.
Electrical broadcasting transponders describing the local road would be the best, but that requires power, so a passive transponder with a description of the immediate roadway which the self-driving car would scan to read would probably work better. Communicate with the car’s programming (AI?) in a way it could more effectively use.
Self-driving cars will have to continue to scan the road ahead as they do now for unexpected or unforeseen obstructions, pedestrians, and no telling what else, but comparing what it scans with what is being told should be there would be better than what it is trying to do now.
Have self-driving cars communicate with each other would help a lot. They could send current velocity, trajectory, and other info so the car’s program knows what is going on better than trying to scan another moving vehicle. Importantly, this should include communicated information from non-self-driving cars, also.
Another significant development for self-driving cars would be communication between traffic devices, such as red lights at intersections. Don’t scan the traffic signal to determine its color. Just have the red light communicate directly with the self-driving car. If this was developed the red light could communicate with self-driving cars at greater distances so that they need not stop at the intersection because the red light could send a message that if the self-driving car would slow or speed up a 5-miles an hour, by the time they got to the intersection, they could pass through without stopping.
Of course, that would not happen at a heavily congested intersection but a traffic signal communicating with all approaching vehicles would change everything, and hopefully reduce a lot of congestion. Once again, new infrastructure may include new – and as yet unthought-of – traffic devices.
The eight hundred pound gorilla in this new technology is self-driving trucks. The development of the interstate highway system changed how we transport goods and supply ourselves, and to not take this into consideration when discussing self-driving vehicles leaves it grievously wanting. I could really see this if self-driving trucks were restricted to the interstate highway system with large parking places at on and off raps. Human drivers would take them to their destination on our local roads.
A test track will all these infrastructure changes mentioned above would go a long way in proving self-driving vehicles are the future of transportation. Development of self-driving vehicles trying to copy human perception of the road has no future – at least not until artificial intelligence catches up with that of humans. And that’s a subject for a whole different posting.