Intelligent Dynamic Replanning for Reduced Airport Operations

Transient delays due to resource constrained reduced operations at one airport can have a significant network-wide impact in a complex airline network. Therefore, an airline needs to dynamically replan its operations given the temporary resource constraints. We consider the problem of prioritizing flight departures post a transient delay. Existing works mainly focus on a full-scale recovery after a significant disruption with constrained optimization approaches. In practice, reduced operations are typically handled manually. We address this gap with a light-weight approach for dynamic replanning. We first build a model for the impact of a delayed flight departure on the additional network-wide delay and missed passenger (PAX) connections along the tail plan of the flight. Next, we formulate and exactly solve an optimization problem of assigning flights to departure slots under a decoupling approximation that limits the interaction between the replanned flights. Finally, we refine the optimizer output with a local search to reflect dependencies among the replanned flights. We evaluate our approach on a real-world dataset of 30 days of tail plans and PAX itineraries. As baselines for evaluation, we use the original schedule (BL1); a greedy scheduling strategy that uses the computed impact model (BL2); and a montecarlo based estimate of the optimal order (BL3). Our approach is scalable with runtime of the order of minutes; and reduces network delay by 11.9% (1.2%, 7.3%); and missed connections by 11% (4.2%, 5.3%) when compared with BL1 (BL2, BL3).