Robert Maxwell wrote:
But lets say in morning traffic we have 5,000 people all going to around the same area about the same time. In theory using a centralized routing system some of the cars would take route X whereas another gorup of the cars would take route Y to maximize road efficiency and reduce travel times.
Some type of centralized coordination would have to occur to make this happen?
Humans are notoriously poor at centrally planning those kinds of systems. Computer science tells us that a self-organizing network of independent agents (that is, individual cars that are aware of their surroundings and even traffic data but not under central control) will often solve routing problems far more efficiently than a central system could.
The Internet provides plenty of real-world relevance here. Routing is decentralized and yet quite efficient and effective.
The internet kind of argues against your point that humans can't do centralization. There are only 13 DNS root servers in the world (called the backbone of the internet)! We don't think about them most of the time because (they're not sexy) and they usually just work. http://answers.yahoo.com/question/in...6182159AATAMFI
This is a centralized-yet-redundant system. I think you're visualizing "centralized" as a single server somewhere with no backups, when in reality, noone would build it that way.
"Centralized" means there is one system or a small set of systems considered authoritative for a given resource. As Lindley
said, DNS only works as well as it does because, despite the very small number of root servers, we all "trust" plenty of servers that aren't
the root ones. We don't all do DNS lookups on VeriSign's servers, we use our ISP, or Google, or even a DNS server we run ourselves (which is, again, pulling from another server up the chain, likely not a root server.)
This approach does have weaknesses, in that a malicious party can inject bad DNS info into the tree, or simple mistakes made at the root level can propagate out quickly and "break" the Internet for everyone. Again, this implicit chain of trust lets the whole system work, but it's also highly vulnerable. It's why DNSSec was developed, which still doesn't have wide penetration.
When talking about high-speed rolling death machines, if you're going to rely on outside data sources that can affect the vehicle's operation, the chain of trust needs to be rock solid and secure.
Instead of going for that, to me it makes much more sense to make the car only accountable for itself, and not under the control of an outside system. The car has to behave as if every other vehicle may do something unexpected and dangerous at any moment, and be prepared to react accordingly. A system in which a central authority tells the car, "don't worry, you can drive within a couple inches of the car in front of you, because we'll make sure it never brakes suddenly," is pretty much begging to be exploited.