Data-Driven Approach Using Machine Learning for Real-Time Flight Path Optimization

Abstract

Airlines traditionally gather weather information before departure to generate flight routes that avoid hazardous weather while minimizing flight time. However, flight crews may have to perform in-flight replanning as weather information can significantly change after departure. This in-flight replanning activity is currently not fully automated, which has the potential to increase crew workload and adversely impact flight safety. The objective of this research is to mitigate some of these issues by developing an automated framework to perform continuous in-flight replanning. The proposed framework relies on three pillars and leverages: supervised machine learning technique to augment existing wind forecasts by providing a higher spatial and temporal granularity, unsupervised machine learning technique to perform short-term predictions of areas with significant convective activity, and graph-based pathfinding algorithm to generate optimized trajectories. The main contribution of this research is to combine these techniques to autonomously and continuously generate trajectories that minimize operating expenditures for airlines. Statistical analyses are performed to demonstrate the applicability and benefits of the proposed framework. Results indicate that optimized trajectories are 2% shorter than actual flight routes in most cases.