Who doesn’t love a day at the theme park? You can go on thrilling roller‒coaster rides, enjoy elaborate shows, have a tasty lunch in between or just relax and take in the scenery. Of course there’s one thing that does spoil the fun a bit: the waiting. For the most popular attractions waiting times of around one hour are not uncommon during peak times, while the ride itself may be over in no more than two or three minutes.
Let’s work towards a basic model for queues in theme parks and other situations in which queues commonly arise. We will assume that the passing rate R(t), that is, the number of people passing the entrance of the attraction per unit time, is given. How many of these will enter the line? This will depend on the popularity of the attraction as well as the current size of the line. The more people are already in the line, the less likely others are to join. We’ll denote the number of people in the line at time t with n(t) and use this ansatz for the rate r(t) at which people join the queue:
The constant a expresses the popularity of the attraction (more specifically, it is the percentage of passers‒by that will use the attraction if no queue is present) and the constant b is a “line repulsion” factor. The stronger visitors are put off by the presence of a line, the higher its value will be. How does the size of the line develop over time given the above function? We assume that the maximum serving rate is c people per unit time. So the rate of change for the number of people in line is (for n(t) ≥ 0):
In case the numerical evaluation returns a value n(t) < 0 (which is obviously nonsense, but a mathematical possibility given our ansatz), we will force n(t) = 0. An interesting variation, into which we will not dive much further though, is to include a time lag. Usually the expected waiting time is displayed to visitors on a screen. The visitors make their decision on joining the line based on this information. However, the displayed waiting time is not updated in real‒time. We have to expect that there’s a certain delay d between actual and displayed length of the line. With this effect included, our equation becomes:
For the numerical solution we will go back to the delay‒free version. We choose one minute as our unit of time. For the passing rate, that is, the people passing by per minute, we set:
R(t) = 0.00046 · t · (720 ‒ t)
We can interpret this function as such: at t = 0 the park opens and the passing rate is zero. It then grows to a maximum of 60 people per minute at t = 360 minutes (or 6 hours) and declines again. At t = 720 minutes (or 12 hours) the park closes and the passing rate is back to zero. We will assume the popularity of the attraction to be:
a = 0.2
So if there’s no line, 20 % of all passers‒by will make use of the attraction. We set the maximum service rate to:
c = 5 people per minute
What about the “line repulsion” factor? Let’s assume that if the line grows to 200 people (given the above service rate this would translate into a waiting time of 40 minutes), the willingness to join the line drops from the initial 20 % to 10 %.
→ b = 0.005
Given all these inputs and the model equation, here’s how the queue develops over time:
It shows that no line will form until around t = 100 minutes into opening the park (at which point the passing rate reaches 29 people per minute). Then the queue size increases roughly linearly for the next several hours until it reaches its maximum value of n = 256 people (waiting time: 51 minutes) at t = 440 minutes. Note that the maximum value of the queue size occurs much later than the maximum value of the passing rate. After reaching a maximum, there’s a sharp decrease in line length until the line ceases to be at around t = 685 minutes. Further simulations show that if you include a delay, there’s no noticeable change as long as the delay is in the order of a few minutes.
(This was an excerpt from my ebook “Mathematical Shenanigans”)