Flight Rescheduling Jo-Anne Loh (jl3963) Anna Ming (axm2000) Dhruv Purushottam (dp2631) IEOR 4405 Production Scheduling (Spring 2016)
Outline Background Situation 1 –Model I –Model II –Comparison Situation 2 –Model III Extensions
Background Airlines and their delays affect the economy –In 2007, passengers experienced a total delay of ~30,000 years with a cost of $16.1B in the US How should flights be effectively rescheduled? –Minimize flight delay –Minimize passenger delay –Minimize revenue loss in emergency situations
Situation 1 Currently airlines reschedule to minimize flight delay. Does this also minimize passenger delay? Goal: Minimize passenger trip delay using Schedule Minimization foR Generalized Operational Logistics (SRMGOL) algorithm Assumptions: –24 hour window –Flights are on a normal distribution –Flights do not depend on previous flights for the aircraft –All passengers have 1 stop over
Situation 1: Data 1)Create initial flight times between 6 airports 2)Add average flight times to create final flight time 3)Create possible itineraries for passengers and generate final arrival time 4)Use percentage late data of each airport to determine which flights are delayed randomly 5)Run algorithms on passengers that miss the next leg of their flight
Situation 1: Data Appendix Average Flight Time(BTS) AirportsORDJFKLAXMIAATLIAH ORD JFK LAX MIA ATL IAH Average Delay Time(BTS) AirportsORDJFKLAXMIAATLIAH ORD JFK LAX MIA ATL IAH Percent Flights Late AirportsORDJFKLAXMIAATLIAH ORD JFK LAX MIA ATL IAH Passenge rs (Random between ) AirportsORDJFKLAXMIAATLIAH ORD JFK LAX MIA ATL IAH
Model I Algorithm The intuitive solution: minimize flight time by putting passenger on the next available flight Prioritizes planes arriving at original schedule
Model II Algorithm Reschedule passenger on the next available flight or hold the connecting flight for a period of time to allow the passenger to make the flight.
Situation 1: Model I vs II ~20,000 passengers, 300 flights Total hours of passenger delay: DatasetModel IModel II
Situation II Aircraft Recovery Problem – unforeseen events disrupt a flight schedule –Ex. In a snowstorm many flights are delayed. Which flights should go first with a limited amount of aircrafts? Goal – reduce losses as much as possible Assumptions: –There are less aircrafts than flights –Flights are full –Relax cancellation constraint
Model III
Model III Data
Model III Results Sample run using AMPL and Gurobi
Extensions Situation I –Include trade-off between passenger delay and airline costs Situation II –Create a dynamic model that can constantly update the unscheduled flights and rerun Combine the two models into a real world situation where there are aircraft shortages and a need to minimize passenger delays