美国数学建模竞赛特等奖论文——2004 B O Developing Improved Algorithms for QuickPass Systems【数学建模】.pdfVIP

Developing Improved Algorithms 319 Developing Improved Algorithms for QuickPass Systems Moorea L. Brega Alejandro L. Cantarero Corry L. Lee University of Colorado at Boulder Boulder, CO Advisor: Bengt Fornberg Summary We model the arrivals at a main attraction of an amusement park by a Poisson process with constant rate; in a more advanced model, we vary the arrival rates throughout the day. The park is open 10 h/day, with a peak arrival time between 2.5 and 6 h of the park opening. We model how a group arriving at the attraction decides whether to enter the normal queue or-to obtain a QuickPass (QP)-a pass to return later for a shorter wait. Their decision is governed by their desire to ride, the length of the normal queue, and the return time for the QP. We explore several models for assigning QPs. The basic model, which gives absolute priority to the QP line, is problematic, since it can bring the normal line to a halt. Our more advanced models avoid this problem by using a boarding ratio-either fixed or dynamically varying-for how many from each line to load onto the ride. We use polynomial regression to predict the behavior of the queues, and we determine a dynamic ratio that minimizes the wait times when the number of QPs that can be assigned per time interval is fixed. Finally, we combine these algorithms and determine dynamically both the boarding ratio and the number of QPs to issue throughout the day. Our advanced models tend to be extrememly robust to small pertubations in starting parameters, as well as to moderate variations in the number of arrivals. We avoid the problems attributed to the current QP system, as long as the ride is not slammed with substantially more guests than