<<< Friday, April 25, 2003 08:29 AM


Friday, April 25, 2003 10:25 AM >>>


Friday,  04/25/03  09:08 AM

The idea: Caravans

I thought about this a bit on my drive home, in traffic.  I was trying to think, what could I do, by myself in my car, to make traffic move along faster.  I think the best thing you can do is tailgate as closely as possible, without actually risking hitting the car in front of you, and track your accelerations and braking as closely as possible to the car in front of you.  A virtual towing rig, essentially.  This minimizes the space your vehicle uses on the road, minimizes air resistance, and maximizes your speed for the cars behind you (by yourself, you can't do anything to speed up the car in front of you). 

I've discussed this with a few friends, and they all have the same reaction.  Paraphrasing, their reaction is "when I get in traffic, I just relax and slow down, and try to drive along smoothly without tailgating".  This is what I do, too.  I don't want to get too close, because then I have to pay attention, and do a lot of stopping and starting.  But this strategy minimizes traffic efficiency!  It allows large gaps to open in front of you, "wasting" freeway space, and all the cars behind you can go no faster than you are going, regardless of the speed at which the cars ahead of you are traveling.

Okay, so what if you built a feature on a car that automatically kept you as close as "safe" to the car in front of you?  {"Safe" would be determined by your current speed and your car's braking ability, and assuming the car in front of you has really good brakes.}  Call this "caravan mode".

How it would work

It seems like this would be technically possible, with a radar or some sort of distance sensor in the front, and a tight feedback loop with your accelerator and brake. 

SO, you merge onto the freeway just like always, get into the lane you want, and hit a switch to enter "caravan mode".  Your car goes into a sort of autopilot, accelerating and braking to keep you as close as possible to the car in front of you.  You are not prevented from accelerating and braking normally (doing so automatically disengages caravan mode, as with cruise control), or from changing lanes, or anything else.  You are not restricted to a particular lane, and you are not dependant upon support from the road or other cars to make this work.  If there is only one car in the whole world with this feature, you still benefit.  For one thing, you don't have to pay attention (!), you won't hit the car in front of you. 

NEXT imagine several cars in a row each with "caravan mode" enabled.  They would travel more efficiently than without caravan mode.  The lead car would set the pace, subject to traffic, road conditions, other traffic, etc.  The other cars would be "towed" along at the same speed, very efficiently.  A long caravan of 100 cards would be WAY better than 100 cards acting independently.  (Interesting to try to model this in a spreadsheet or something.)

Some possible problems and solutions:

  1. What if the lead car travels too fast for one of the other cars?  The solution is simply to have an adjustment for the maximum speed you're willing to go.  Just like a cruise control, essentially.  If the car in front goes faster than this, you fall back and leave caravan mode.  As soon as you get close enough to a car in front, you reenter caravan mode.
  2. What if you follow too close?  As with the maximum speed, there would be an adjustment for this - turn the knob to select.  Some people are more comfortable being further back - the driver decides.  Leaving a bigger gap is safer and less aggresive, but also less efficient.
  3. How do other cars merge into a caravan?  This is a tough problem, because each car in the caravan is automatically trying to minimize the gap to the next car.  In the early days of this feature few cars would have it, so caravans of more than a few cars in a row would be rare and merging would not be too hard.  If the feature is successful and many cars have it, long caravans would form, and merging would become difficult.
    A couple of possibilities for this:
    1. The human driver could notice that a car wants to merge (blinker, etc.) and touch their brake to slow down.  This would open a gap, and the car could merge in.  Once the merge is complete, the driver re-engages caravan mode, and life goes on.
    2. If the human driver doesn't slow down, and a car "sticks their nose in" anyway, so be it; the caravan sensor notices the new car, slows down (to maintain a safe gap), and the newcomer joins the caravan.  If the new car doesn't have caravan technology it becomes a new lead car.
    3. If the car merging in was equipped with the caravan feature, it could electronically signal the cars in the caravan.  This would cause the nearest car to slow down automatically to make room.  In the coolest case the merge would happen automatically with no human intervention required by the driver making room, but perhaps s/he has to reengage caravan mode manually after the car has merged.  {One pleasant attribute of this solution is that it creates additional incentive for people to add caravan mode to their car, so they could merge into existing caravans...}
  4. What happens when the lead car exits the highway?  Nothing!  This is a not a system for steering, it is only a system for controlling speed.  The lead car changes lanes, exits the highway, etc.; so the second car becomes a new lead car.  It will speed up to the maximum level set in its cruise control until it gets close to another car, at which point the car in front becomes the new lead car.  {An interesting attribute of this approach is that lead cars will most likely not have this technology.  So cars which have caravan technology will "draft" cars which don't...}


This solution has some pretty properties (W=UH):

  • No road modifications are required.
  • It appears technically feasible at a reasonable cost per car.
  • Cars with the feature can coexist with cars that don't have the feature.
  • Cars with the feature benefit even if very few other cars have the feature, but the more cars have the feature, the valuable the feature becomes.  (Leading to a classic "network effect".)
  • It is a safety feature.  Really!  You will not be able to hit anything in front of you while in caravan mode...  At the very least it is safer than cruise control, which is installed in just about every modern car.

More complexity

Here's a v2 feature:  Cars in a caravan “communicate” with each other, via a wireless network (802.11, anyone?).

The primary benefit of the caravan is that it eliminates the “lag” in human reaction from the time the car ahead of you changes speed to the time that you do.  However, there is another lag caused by the inherent momentum of the car.  Even if the caravan electronics instantly detected a change of speed in the car ahead and instantly changes your car’s acceleration or braking, your speed would not change instantly.  Sure, eliminating the human lag is great – by far the main part of the overall lag – but eliminating the “acceleration lag” would be even better.

If the car ahead could signal that it is braking or accelerating before it actually does so, your car could anticipate the change and compensate automatically.  The next level of improvement would be a signal from the car ahead of the car ahead – anticipating adjustments by the car ahead before it actually makes them.  Perhaps the added complexity of this solution would not be worth the limited benefit.

Another idea – have predictive software in the car which anticipates acceleration changes.  This could be a learning algorithm, so your car gradually becomes more familiar with the behavior of the car ahead.  This has the advantage of simplicity (no communication with other cars required) and can also be done in “what if” mode, without actually changing acceleration, to see if it would help.

Even more complexity

Here's a v3 feature:  Regional networks.  Cars with caravan technology could "subscribe" to a service which helps them locate caravans to join.  This network could also signal weather, accident information, and traffic conditions.  {In addition to being useful, this is also a recurring revenue opportunity...}

A little off subject:  I have a little GPS map unit in my car, I love it.  The one thing it lacks which would make it totally great is information about traffic conditions.  It will pick a route it feels is optimal but which I know involves the Sepulveda Pass at 8:00 in the morning.  I’ve often thought it would be cool if all people who had such a unit were connected through a two-way network, probably piggybacked on the two-way paging networks.  Each car would contribute its own traffic conditions (just its location and speed, otherwise anonymous) in exchange for realtime traffic information from other cars in the vicinity.  There are enough people driving around with GPS units that a reasonable picture of realtime traffic on major streets would result.  Something I would gladly pay monthly for…

Your comments and suggestions are eagerly solicited...

[ This article was originally an email thread; Greg Crandall, Chris Prajzner, Kevin Schantz, Nick DeNicholas, Russell Bailinson, and Ramon Kurkchubasche each contributed to these ideas. ]

[ Later: the Caravan Fallacy... ]

[ Even Later: Caravans Revisited - the future is here! ]

[ Much Later: Caravans cont. - back to the future ]