美国数学建模竞赛特等奖论文——2006 B O Application of Min-Cost Flow to Airline Accessibility Services.【数学建模】.pdfVIP

Application ofMin-CostFlow 367 Application of Min-Cost Flow to Airline Accessibility Services Dan Gulotta Daniel Kane Andrew Spann Massachusetts Institute of Technology Cambridge, MA Advisor: Martin Z. Bazant Summary We formulate the problem as a network flow in which vertices are the loca- tions of escorts and wheelchair passengers. Edges have costs that are functions of time and related to delays in servicing passengers. Escorts flow along the edges as they proceed through the day. The network is dynamically updated as arrivals of wheelchair passengers are announced. We solve this min-cost flow problem using network flow techniques such as price functions and the repeated use of Dijkstras algorithm. Our algorithm runs in an efficient polynomial time. We prove a theorem stating that to find a no-delay solution (if one exists), we require advance notice of passenger arrivals only equal to the time to traverse the farthest two points in the airport. We run our algorithm on three simulated airport terminals of different sizes: linear (small), Logan A (medium), and OHare 3 (large). In each, our algorithm performs much better than the greedy send-dosest-escort algorithm and re- quires fewer escorts to ensure that all passengers are served. The average customer wait time under our algorithm with a 1-hour ad- vance notice is virtually the same as in the full-knowledge optimal solution. Passengers giving only 5-min notice can be served with only minimal delays. We define two levels of service, Adequate and Good. The number of escorts for each level scales linearly with the number of passengers. One hour of advance notice is more than enough. Epsilon Airlines can make major improvements by using our algorithm instead of send-closest- escort ; it should hire a number of escorts somewhere between the