An Integer Programming approach to Dynamic Airspace Configuration

Given the relevance of aviation as transportation network and its remarkable economic impact, the air traffic demand is bound to increase. High traffic density in a given airspace region can cause safety issues and difficulties in air traffic control task, so it is necessary to efficiently organise the airspace structure to avoid under and over-loaded areas of the airspace. To this end, we model the airspace by means of airspace blocks (3D portions of the airspace) and sectors (3D connected unions of airspace blocks). A capacity is associated to each sector, limiting the maximum number of flights that can enter said sector in a time interval. An airspace configuration is the partition of airspace into sectors in such a way as to accommodate traffic as efficiently as possible. Given a pre-determined set of configurations with related capacities and the dynamic air traffic demand in a time horizon, we aim to determine a sequence of configurations (configuration plan) that optimally meets the demand. The sequence must also satisfy operational restrictions that smooth the configuration dynamics, as to avoid, e.g., too frequent switching between configurations. The problem is known as Dynamic Airspace Configuration and is mostly faced by means of heuristic approaches. We propose an Integer Linear Programming model that provides a configuration plan for a given timeframe that minimizes the traffic overload with respect to the capacity, and we test it on five days of historical data over the Madrid ACC. We compare the results of different time discretizations and the impact of traffic increment on the traffic overload of optimal configuration sequences.