16 thoughts on “Traffic Jams”

  1. “Another thing they’ll do is to all start moving when the light changes, so we won’t wait until the car in front of us moves.”

    Except that the reason we wait is so that we have sufficient braking distance in case of the need for a quick stop. (and the fact that some drivers are texting 😉 )

    Your AI car cannot move until safe braking distance between you and the AI car ahead has been achieved.

    Your AI car will have to wait for a few instants to discern the speed of the AI car ahead so that it can know when to begin moving – so there will be some delay before moving.

    And the car ahead will not be going at an instantaneously fixed speed but will be accelerating up to some speed. So the AI car will have to accelerate more slowly in order to maintain safe braking distance, since the braking distance increases with increased speed.

    And all of this has a ripple effect back down the line. Same as with humans. Maybe the effects will be smaller but they won’ t be zero.

    1. Indeed, but safe braking distance is shorter with the faster reaction time, and the car can start accelerating while the driver is pre-occupied texting. In practice, I noticed the slow acceleration is more a nod to fuel efficiency rather than safe braking distance. The car seems to notice a vehicle moving away from it (faster velocity/acceleration than itself) and ignore it.

  2. Birds can figure this out, but not humans. If we can’t build AIs smarter than ducks, best not have them at all. Seriously, the problem isn’t braking distance or reaction time, it’s that most people don’t start to move until the person in front of them has completely driven away. I’ve always wondered if this was mainly due to abdication of responsibilty. The person at the front of the line is supposed to watch the light. Everyone else just watches the back of the car in front of them. I’ve been driving for 50 years, and driving habits are the same as they always were.

  3. “Everyone else just watches the back of the car in front of them.”

    Everyone else just waits until they notice the car in front of them is moving before they stop texting and start driving. This gets really annoying at short left turn lights, because of course the oblivious idiot keeps on going even though the signal has gone red. (Sorta like how the guy dawdling along slower than the limit finally speeds up as the light turns yellow and plows on through even though it turned red before he got there.)

    One thing I do like are the timed crosswalk lights. They give a useful indication of how soon I’ll see the yellow, and can plan accordingly on what I need to do.

  4. Bill Burr and George Carlin have both independently come up with the same solution. Kill more people and produce fewer of them.

    I say get rid of smart weapons that have too low a body count.

    But seriously, population density above what the infrastructure supports is a real problem. Why is it ok for a single mom to have a dozen kids with multiple baby daddies that responsible people have to pay for?

  5. I remember doing this problem as an undergrad at Purdue. It was an applied engineering course, and the real test was whether one could remember to take the first and second derivatives of a curve fit, and find out whether the latter was negative or positive when the former was zero. So this is hardly a new “finding.”

    The much, much more difficult (and as yet unsolved) problem is how to distinguish between traffic jams and traffic jellies.

    1. In 1904, the first subway in NYC had a track fire causing the jelly to jam! Commuters have been late for work ever since!

      I always thought Phoenix would be a good place for subways (and underground housing with a year round 70 degree temperature) but was told the water table wouldn’t allow it… which makes no sense to me since some subways in NY go under the river between Manhattan and Brooklyn every day.

  6. Interesting. The article was based on the results of a simulation of traffic movement. Usually Rand is skeptical of simulations, at least when it comes to “global warming.”

    Given I didn’t read about any validation of the model, I’m taking this one with a large grain of salt.

      1. Any simulation can be completely bogus either by leaving out parameters or making biased assumption. Then there’s the butterfly effect which is almost always present.

    1. Simulations can be very useful if done right. But you get problems when, like climate simulations, assumptions are made and the results are not subject to validation.

  7. Smart car technology will involve more than seld-driving capabilities. The cars will be networked to one another (and I sincerely hope that they’re designing in cyber security from the beginning instead of it being an afterthought). When the light turns green, the cars can pull out together. The technology will reduce tge safe separation distance between cars which will increase highway capacity. It will reduce the ripple effects that happen when cars slow for no good reason. The cars will be programmed to obey the law which will eliminate a source of revenue from local governments that depend on speed traps and predatory enforcement. I’m 60. I hope this technology becomes widespread and affordable before I’m too old to drive myself anymore.

    1. A local area network to coordinate AI cars may be useful. But trusting messages from other cars and traffic lights too far is asking for trouble the moment bad data in injected into the network, whether because of a bug or deliberately.

Comments are closed.