Exactly! That is the type of efficient behavior that I would expect a self-driving car to have. But what if I am in a hurry and I pull over onto the shoulder to pass a line of cars? Human drivers in that line would be sure to exhibit any number of asocial responses, but self-driving cars could be counted on to avoid an accident and make way without any fuss. The best deterrent would be to document the behavior and perpetrator via the sensors and send a message to the local traffic control. This would be the modern equivalent of a citizen's arrest.
Sure, smart by itself has great value. But V2x communications greatly strengthens the value proposition. Straightforward example is how real-time knowledge of traffic conditions would allow for trip rerouting, thus improving efficiency, saving time and reducing driver/passenger frustration.
True. But the most efficient will be that both lanes will have about equal length queues, and cars will very rapidly interleave at the choke point. Much more rapidly than they could when driven by humans.
This shortens the length (in feet, not time) of the queue, by making effective use of the real estate of both lanes, and it will also end up being "fair." A shorter queue creates less blockage of alternative routes upstream, and as to fairness, the guy in the lane that's blocked will be in queue as long as cars in the other lane.
This is why I talked about the bad behavior that the earlier post described rather than just "bad" behavior in general. When most people talk about bad driving they are generally talking about people that drive differently than they do. If/when we automate driving there are sure to be people that will be extremely annoyed when their car does something different than they would have done. Example: Two lanes, the right lane is ending. Traffic line is longer on the left. Do you hit the right and merge in ahead of others? That annoys the crap out of me, but network theory says that that is the best choice (it avoids blocking up traffic in intersections behind you). If I do that I am being rude. If my car does that it is being efficient.
Actually, when I think of bad driver, I think of the idiot who is texting while driving, or who falls asleep at traffic lights (never mind while driving!). In any event, these guys, as well as the overly agressive, will have less of a negative impact on traffic.
It's not just "safety." It's also about making efficient use of the roads. If you're at an intersection, and the genius in front of you forgets what it means when the light turns green, the result is that this intersection will saturate more quickly. That catatonic driver makes it so fewer cars can make it during the traffic light's cycle, and that promotes traffic jams.
Same with drivers who don't know where they're going, so they decide to do their introspective soul searching right in front of a green traffic signal. If the car would either drive itself, or at least tell this guy where to turn (well within today's state of the art), again, the roadways would be used more efficiently.
A car would be fairly useless if there were no roads. So in fact, that aspect of Metcalfe's law has long been with cars.
I fully agree with KB's points, though. Connected cars can appear on the scene very quickly, even if fully autonomous operation will be introduced gradually. And vehicle to vehicle and to infrastructure comms are imperative for fully autonomous operation, for fairly obvious reasons. Think how you drive a car manually. You try to decipher what that guy in front of you is thinking, and you look at the road. Those actions have to be automated, not ignored. If the car in front of you puts on a turn signal, it's because this intention to turn matters to others, right? It's supposed to make drivers approaching and follwoing this guy do something specific, different from how they would behave if no turn signal were present. Well, there you go.
Another point. I remember very well, back in the 1960s, that often when flying across the Atlantic, planes would get stuck in an endless holding patterns before being able to land at JFK Airport. Why did that problem subside? Mainly because of scheduling improvements. Airlines don't ignore every other airline's scheduled arrival times. Same improvement can happen in cars, wrt vehicle to infrastructure comms, gradually. Those with smarter cars will get stuck in fewer traffic jams. Either because they'll figure a better route before leaving, and/or because they'll leave at a more optimal time.
The problem that you describe will get worse before it gets better. The bad driver behavior that you describe depends on intimidation of other drivers, forcing them to accomodate the bad driver. That behavior increases as the accomodation increases, since the reward also increases (bullies always escalate if they are not pushed back). How do you think those drivers will react to cars that can be relied on to give way? I predict that it will become a sport. Self-driving cars will be seen as weak and turned into targets for intimidation.
Junko said: "While I agree with you every point you raised here, this actually raises another question. Then, do we even need V2V or V2I? If every car gets so smart after all, what's the point of V2V and V2I?"
Well, for V2V, one scenario I can envisage is for a car to follow another car automatically. For example, I often have to follow someone else when driving in foreign countries, it would be nice if my car could do that automatically just by communicating with the other car. This needs initial secure identification and authentication and then data to be sent forwards and backwards between the two cars.
For V2I, I can think of the automatic driving scenario as well e.g. to get updates on the state of the roads/traffic for rerouting purposes. Another scenario is for the car to connect to an intelligent home hub e.g. to check on the stock of groceries and perhaps suggest stopping by the shops on the way back home.
While the value of state-of-the-art cars with V2V, V2I etc. is maximised with increasing numbers of other state-of-the-art cars o the road, it does not make them useless nonetheless... A smart car for instance is better equipped at a avoiding obstacles and crashes, optimising fuel consumption, keeping the people onboard connected with the rest of the world etc. etc.
While I agree with you every point you raised here, this actually raises another question. Then, do we even need V2V or V2I? If every car gets so smart after all, what's the point of V2V and V2I?
Blog Doing Math in FPGAs Tom Burke 23 comments For a recent project, I explored doing "real" (that is, non-integer) math on a Spartan 3 FPGA. FPGAs, by their nature, do integer math. That is, there's no floating-point ...