Research on delay propagation mechanism of air traffic control system based on causal inference

Abstract

In the air traffic control (ATC) system, delays are tightly coupled among airports. Understanding the interactions of these delays is a prerequisite for accurate prediction. However, most researches in this field depend on correlation methods, and few studies apply causal discovery methods. The latest approach uses the cause-and-effect diagram discovery algorithm PCMCI to mine the causal dependencies between time series. In the application context of the ATC system, not all delays have the propagation effect because of a series of optimization measures by airports and airlines. This paper proposes an extreme event measurement mechanism to solve the above problem by setting the airport thresholds adaptively through a complete search of the parameter space. The improved method allows for a more accurate exploration of the time lag delay propagation relationship between airports. The experimental simulation results show that the algorithm in this paper can extract time lag causality in nonlinear, multidimensional time series considering the ATC system characteristics. In addition, we use the causality of airport delays to construct an extreme delay propagation network (EDPN), use complex network theory and related metrics to analyze the network at the system level, and propose a new critical airport identification metric. In the case study of the U.S. aviation system, we find that small airports have a weak ability to mitigate extreme delays and are more likely to propagate delays to other airports in addition to their own impacts; Larger airports have strong mitigation and robustness to extreme delays but suffer from overload operations. These findings are helpful to explain the interactions of delay processes, and can provide specific theoretical support in improving system operation efficiency and developing delay propagation mitigation measures.