The Effect of the Planning Horizon and the Freezing Time on Take-off Sequencing

In this paper we consider the problem faced by runway controllers at London Heathrow Airport as they attempt to determine the best order for aircraft to take off. The order in which aircraft take off can have a large effect upon the throughput of the runway and the consequent delay for aircraft. Although two runways are available for use at Heathrow, in order to control noise for residents on the flight paths, only one is used for departures at any time. The other is used for arrivals. As this is one of the busiest international airports in the world, and the runway is the bottleneck for the departure system, it is important to attain high runway throughput. A runway controller re-arranges the aircraft for take-off within holding points at the ends of the runways, currently performing this task manually. The decision about the take-off order has to be made with very limited decision time, precipitating investigations into the development of an on-line decision support system to aid in this task. We have developed a model for such a decision support system and a simulation of the departure system, and discuss these in this paper. Our experiments predict that our system can provide suggestions fast enough to be of use in practice, while also being able to consider more aircraft than a real runway controller can. Thus it can obtain consequent benefits from highlighting potential problems before they occur. In order to maximise the potential benefits of a decision support system, it is important to understand how the decisions are affected by the planning horizon, how far ahead the system considers aircraft. From this information, we can better understand the inputs that the system would need to be able to fulfil its role. The position of an aircraft in the take-off order has to be frozen at some point prior to takeoff. Our investigations revealed a trade-off between the time at which the take-off order was frozen and the planning horizon necessary to get the best results from the system. This paper considers this trade-off, evaluates where the planning horizon needs to be, and shows the relationship with the point at which the take-off schedule is frozen. We present results which show that there is a substantial delay benefit from including taxiing aircraft. Furthermore, we show that, with a schedule frozen for two minutes before take-off, the vast majority of the benefit can be gained from a knowledge of only those aircraft which will arrive at the holding point within the next eight minutes, and that aircraft can be ignored until they push-back from their stands. Finally, the results also show how the planning horizon must increase if the time for which the schedule is frozen increases.