Network-wide robust and resilient metaheuristic trajectory optimization under thunderstorm disruptions

Abstract

Network-wide robust and resilient trajectory planning is realized after the uncertainty propagations at trajectory and ATM levels. The inputs are the 4D trajectories with uncertainty and the delays applied to trajectories for network resiliency. The delays only shift the trajectories in time. The output is a set of algorithmic solutions for optimal trajectory selection under high complexity situations. According to the START concept, a proposed rerouting and/or rescheduling solution of the user-preferred flight plan is proposed to improve the resiliency and robustness of overall planning. The optimization process is realized using the simulated annealing metaheuristic to find the optimal rerouting and delays for each flight. The objective function of this optimization problem is a complexity metric function, based on Linear Dynamical System. This metric can consider uncertainty in the 4D trajectories. However, the computation of such metric requires extensive computation time. We proposed GPU-based concept to speed up the metric computation. We have found that the proposed GPU-based concept can potentially provide the desired performance and prove the computational viability of the START project. Nevertheless, our findings are not uniformly positive, as the reliance on single-precision arithmetic (on which current GPUs provide substantially higher throughput) seems to have proved more problematic than our previous expectation. The global air traffic complexity is reduced by a factor of six hundred from around 120 to 0.2. It corresponds to a better organization of the traffic. In fact, the complexity is mainly due to very few flights. The complexity reduction decreases the potential number of conflicts, because there are less converging air traffic situations.