An Approximate Dynamic Programming Approach to Aircraft Allocation Incorporating Passenger Demand Models

Passenger demand for airline transportation is a function of numerous factors. The way that airlines provide air transportation to the passengers determines some of these factors; this includes factors like frequency of service, number of stops on an itinerary, size of the aircraft, and ticket price. As potential travelers observe these factors, they make decisions about future travel. This paper presents an approach in which an allocation problem represents how an airline offers service to meet passenger demand at a given time step. The results of this allocation become inputs to an econometric model that predicts future air travel demand as function of the current observed service. Establishing a feedback from the demand model to a subsequent allocation model and using concepts from dynamic programming, making allocation decisions that could maximize or minimize an airline objective function (e.g. maximize profit, minimize cost) over a finite time horizon becomes possible. An example using this dynamic programming approach for a very simplified route network demonstrates how an airline might consider which type of new aircraft would lead to the maximum aggregate profit for the airline over a finite time horizon.