Week 11 – Obtaining Results
Hey all, busy week! This week I focused on obtaining all of the remaining results and testing out our reservation system. So, without further or do, let’s begin.
As shown last week, our Type A reservation system had these metrics:
Total Waiting Time: 15921 Ticks ~= 2653.503 minutes
And 7.27 minute average waiting time
This represents the status quo and a normal undisturbed area of Basis Scottsdale.
And for this week, I focused on obtaining Type B and Type C:
Type B is quite tricky. The idea is to split the Queue into multiple chunks (2,3,4,5, etc.) to reduce the total waiting time, in expense for the total time horizon. This would represent an “ideal” scenario wherein all agents follow the same, predictable waiting time with minimal delay.
(Fig. 1. Type B control, where I split it into two queues (schedules) and have a 10 minute buffer time for late/early arrivals).
After many arduous 2AM nights, the result was quite great:
Total Waiting Time: 12239.4 Ticks ~= 2039.9 minutes
Or 3.98 minute average waiting time
An almost 2x improvement from Type A control. But this is impractical since we’re using both lanes as a schedule. Instead, I wanted to make the users choose – so that’s Type C.
Type C is even harder. For this we need to calculate the equilibrium point at which users will choose one lane over another, giving them full autonomy over their decision-making. From what I found, the ideal point to change is about at 108 cars, meaning if there are over 108 cars in the non-scheduled lane, the cars will pivot to the scheduled lane and vice versa.
Then, we calculate the waiting times for each and average them out. The results are equally as pleasing:
| (in minutes) | Type A | Type B | Type C |
| Total Waiting | 2653.503 | 2039.893 | 2602 |
| Average Waiting | 7.268183 | 3.984167 | 5.589044 |
| Time Horizon | 60 | 85.33333 | 60 |
As shown in this table, Type C is the most feasible as it doesn’t change the Time Horizon, nor does it require everyone to cooperate. Type B, while the best, does have issues in that regard. Both scheduled systems drastically improved the average waiting time, and thus could save hundreds of hours of waiting time across the nation each day.
After obtaining the results, I’ve ran the charts and am currently writing a paper. Will update once more on the final progress! Thank you for reading!