This is a short boarding time simulator inspired by CGP Greys video on better boarding methods.
Python 3.x should include all the packages required for the boarding time simulator but see the dependencies for visualization tools. To run a simulations, execute
from airplane import Airplane
from simulation import Simulation
from queues import FrontToBack
airplane = Airplane(number_of_rows=30, seats_per_row=6)
sim = Simulation(airplane, FrontToBack, max_iter=1000)
sim.run()sim.run() returns a dictionary with the number of simulation steps, number of stops (i.e. when a passengers sits down), number of seat shuffles, and passengers seated. I decided to calculate the time for boarding as time_for_boarding = results['number of stops']*TIME_PER_STOP + results['number of seat shuffles']*TIME_PER_SHUFFLE with 1 minute for each stop and an additional 0.25 minutes when seat shuffles are required.
For visualization I used the following packages.
- matplotlib
- pandas
- seaborn
In èxample_simulations.ipynb I collated example simulations for the boarding methods discussed in CGP Grey's video: front-to-back, back-to-front, random, window-middle-aisle, modified Steffen, and perfect Steffen.
For the examples here I used a single-aisle plain with 30 rows and 6 seats per row (window-middle-aisle on either side). I did not include a first-class in the plane for simplicity. Each simulation is run 100 times and the results are plotted using
In Front-to-Back the passengers start boarding the plane with the first row and fill towards the back. Front-to-Back is probably the worst boarding method because it prevents parallel seating because only the person at the front of the line has reached their seat at any given point. In this example, there is on boarding group per row. Each row boards in order but passengers within each row are randomized.
Back-to-Front is the opposite of Front-to-Back. The last row starts boarding and the plane is filed according to decreasing rows. As with Front-to-Back, rows board in order but passengers are randomized within each row.
For random seating all passengers are randomized and there are no boarding groups.
In Window-Middle-Aisle passengers are organized into 6 boarding groups starting with window-seats on either side, then middle-seats on both side, and lastly aisle-seats on both sides. Within each group, passegers are randomized so there is no order in terms of the row number but seat-shuffles are eliminated.
Steffen is an idealized boarding approach that is theoretically most efficient (see Steffen perfect). However, to adapt the Steffen method to real-world boarding passengers are sorted into 4 boarding groups: Alternating rows on alternating sides. Within each group, passengers are again randomized.
As mentioend above, Steffen is an ideal boarding method that is theoretically most efficient. Passengers are perfectly ordered both back-to-front and window-middle-aisle. In the method mentioned by CGP Grey, each group is separated into alternating rows (e.g. first odd, then even rows) resulting in 12 boarding groups. However, I decided to omit the even-odd split to simplify the simulation steps resulting in 6 boarding groups (same as in Window-Middle_Aisle but ordered from back to front).
In the video, modified Steffen is depicted as though people are sorted and board back to front. I decided to board people in random order which seems closer to the originally proposed method to address "human imperfections" (wiki).
I also ignored several features of real-world boarding that may decrease or increase more realistic boarding times:
- no groups: each passenger boards independently. In the real world, friends, families, colleagues would frequently board the plane together.
- moving through the plane takes no time. While the number of steps is given, I decided to ignore the time required to move in the analysis. This would especially effect the magnitude in difference between perfect Steffen and the other methods
- I also ignored first-class and early boarding to simplify the simulation