Algorithm Challenge: Burning Man Vehicle Exodus

Algorithm Challenge: Burning Man Vehicle Exodus 273

Posted by Soulskill on Friday April 04, 2014 @10:53AM from the in-the-name-of-efficiency dept.

Slashdot contributor Bennett Haselton writes: "A year ago, getting ready for Burning Man, I read that the cars in the exit line sometimes have to wait in the sun for hours to get out. I came up with an algorithm that I thought would alleviate the problem. Do you think it would work? If not, why not? Or can you think of a better one?" Read on for the rest of Bennett's thoughts.

As part of my advance preparation for going to Burning Man in 2013, I read on the official site that the car lines to get out of Burning Man often take five hours to get through. Scroll a bit further down and you can find, asked and answered, the question that I thought of after reading about the five-hour waits, and it's worth quoting the whole thing:

Q. You should set up a system where people can register for a departure time and give them an "express" lane (or some version of a priority/regulated system). Those who miss their window or don't register would have to wait longer.
A. This suggestion has made its way to us every year for many, many years now. And on the surface it looks very attractive. But, as is usually the case, the devil is in the details. Here are the primary reasons we have not implemented a reservation-based Exodus system:

Such a system takes a lot of people power (e.g., people to verify departure times, people to direct traffic, people to enforce the system) and a lot of resources (e.g. a registration system, building secure lanes for 5 miles of Gate Road that would prevent people from jumping into the wrong section)...more than we currently have.
Verifying registration would require slowing traffic before Gate Road, which will in turn slow down the rate at which people can get onto Gate Road. Without a significant redesign, traffic inside BRC could become gridlocked.
One thing we have learned about Burning Man is people rarely stick to their intended timeline. Camp clean up took longer than planned, you stayed up really late the night before, it took a while to track down your passengers, you couldn't find your car keys, you just had to visit the ashes of the Man one more time, or myriad other possibilities that are so very common to the Burning Man experience. To get 50,000 people to stick to a specific window of time may very well be the most difficult part of this idea to solve.
Another thing our Gate experience tells us is that verifying Exodus registrations and enforcing 'rules' will not be a cut-and-dried process. We will no doubt hear many stories (traffic to get from my camp at 2:00 was worse than I thought, but I really did leave in time! My camp-mate burned my registration slip in an offering to the Man but this really is my time window! I have a flight that leaves in a few hours, please I need to get out faster!). Each vehicle that pleads their case in turn holds up traffic for everyone else, and this ultimately will cause significant inefficiencies in the system.
Remember how we said this type of system would require a lot more people power? Despite our calls for help from the community, we continue to struggle to find enough people to manage the bare basics of Exodus (e.g. highway flaggers). We understand that most people are tired by the end of the event, and many need to get home. However, in order for us to continue to evolve the Exodus process, we need YOUR help. We need volunteers to help run all parts of this process. Everything that happens in BRC is created entirely by its citizens, including Exodus.
Some of the above issues could be overcome, but taken all together a system like this in an environment like Burning Man would be complex and expensive to implement and considerably more difficult to run efficiently.

Bennett again. So I thought about this some more and wondered about a different idea: My question: Why not have a priority exodus line set aside for vehicles who leave during a designated time slot, based on the last digit of their license plate? So for example halfway through Burning Man, a random number or letter would be selected by the organizers — say, "T." During daylight hours on the last day, a priority exit lane is set up where from 6:00-6:30 AM, only vehicles with license plates ending in "T" can exit. Then from 6:30-7:00 AM, only vehicles with license plates ending in "U" can exit. And so on, until you've cycled through all the letters and numbers. (The initial letter in the cycle — in this case, T — would have to be selected after the event starts, to prevent people from gaming the system in advance, by bringing in vehicles with plates deliberately chosen to get an early exit time.) And then you have a second, longer line for everybody else who doesn't want to leave in their designated time slot.

This has a number of desirable features:

It avoids most of the problems described in the FAQ — you don't have to "create" a registration system, or stop cars in order to verify their registered departure time. All you need are observers for the priority exit lane watching to see that the cars in that lane have the correct last digit of their license plate. (Since all exiting cars are passing through the same bottleneck, you only really need one or two observing at a time to glance at license plates.) And if an observer spots a cheater, they don't have to throw their body in front of the vehicle, just radio ahead to tell someone further down the road that there's an unauthorized car in the priority exit line.
It's difficult to cheat. You could try to hack the system by bringing multiple sets of license plates to Burning Man and then, after the departure times have been announced, putting the earliest-departure license plate on your car. However, apart from the fact that this is illegal (which never stopped certain recreational activities at Burning Man, after all), there would be diminishing returns from loading up on too many extra license plates. If you want a guaranteed exit in the first 9 hours, then out of 36 sequential time slots, you'd only need 4 different license plates to guarantee an exit in one of the first 9 slots. But if you wanted a guaranteed exit in the first 3 hours, then you would need 12 different license plates, and so on.
Most importantly, and this is the whole point, would reduce the amount of time waiting in the exit line, for drivers that opted to use this system. Under the existing system, with a single queue that anyone can enter at any time, the queue grows to a length at which the inconvenience of the long wait is just barely outweighed by the desirability of getting out (an equilibrium which apparently sometimes causes the lines to grow to up to five hours). By dividing the population into segments by last digit of license number, those drivers are only queueing up with 1/36th of the rest of the population, and so can expect a faster exit time.

In the theory of queueing, if a population is sufficiently large, then when users are queueing for a desirable resource, the queue will grow until the cost of waiting in the queue is just barely outweighed by the benefits of the resource at the end of it. (Steven Landsburg explains in the opening chapter of The Armchair Economist that if a sufficiently large town opens a free aquarium, the line to get in will grow to the point where the inconvenience of the line exactly cancels out the benefits of the visit, so the benefit to the citizens' lives will be exactly zero.) Interestingly, this means that for the Burning Man exit queue, if you simply divide the queueing population in half — say, by allowing cars with even license plates to exit in the morning, and cars with odd license plates to exit in the afternoon — then you won't accomplish anything, because each half-size population will probably still be large enough that the queue grows to the point where the convenience of getting out just barely outweighs the inconvenience of waiting in line. It's not merely that dividing the population in half wouldn't accomplish as much as dividing it into 1/36th slices; it's that dividing the population in half would accomplish nothing at all. To make the queue shorter, you have to divide the population into sufficiently small slices that there is no longer a large enough population in each slice, to make the queue swell to the point of convenience-cancelling equilibrium. The simplest way I can think of to do that would be to split up the car population into 1/36th by last license plate digit.

It's important to note this does not actually increase the rate at which drivers can exit from Burning Man, which is actually a limit set by the Bureau of Land Management at 1,000 cars per hour. No algorithm can get around that limit. The algorithm only aims to reduce the amount of time that cars spend waiting in line to get out (in the hot sun, some with broken air conditioners). If you want to use the prioritized queue but you know that your time slot won't come around until 2 PM, you can spend the time until then exploring what's left of Burning Man, learning and making new friends, instead of getting in line at 10 AM just to get out by 2.

In any case, this isn't my problem, since I took the Burner Express bus in and out of Burning Man and would plan on doing it again. But while I was preparing last year, I went ahead and posted the question to ePlaya, the Burning Man message boards ("playa" being another word for dry lake and the nickname for the physical location of Burning Man). Some of the respondents were convinced that "Bennett Haselton" was an elaborate troll (you guys would get along), although I mostly got people saying, "The organizers have had years of experience doing this, why not wait and see it in person before trying to 'solve' it." Well, I was kind of asking for it, admitting that I had never been to Burning Man before, posting in a forum frequented by grizzled veterans, claiming that from my ivory tower on high, I had divined a solution to a problem that others had been working on for decades. (Of course, none of these are valid reasons why the idea is wrong.)

But anyway, I took the advice in the replies: as I was riding out of Burning Man in the Burner Express bus, I glanced out the window as we passed a mile of non-moving cars waiting to get out. I still don't know what I was supposed to see that would illustrate why the license plate prioritization system would a bad idea. What do you think? Or do you have a different idea?

Then again, maybe it doesn't matter how objectively "good" an idea is, if change is just plain hard. In another thread that I started after Burning Man was over, I said that the porta-potties seemed to work fine but that the dispensers next to the porta-potties, mounted on wooden stakes stuck into the ground, were almost always empty. They could easily attach more dispensers to the posts, or set up more posts (as long as the maintenance company kept replenishing the dispensers with the same frequency), at a cost that would be almost nothing relative to the cost of maintaining the porta-potties in the first place. Even that suggestion was met with fair bit of snark, although eventually someone gave me the email address where I could send feedback like that to the Burning Man organizers. So I sent the hand sanitizer suggestion to the feedback address, but don't hold your breath (except in the porta-potties).

This discussion has been archived. No new comments can be posted.