Pilot mental workload analysis in the A320 traffic pattern based on HRV features.

Abstract

Pilot mental workload is a critical factor influencing flight safety, particularly during dynamic flight phases with high cognitive demands such as takeoff and landing. This study evaluates pilot workload across different flight phases (takeoff, climb, cruise, descent, and landing) using HRV (heart rate variability) features and machine learning methods. Heart rate data were collected through simulated A320 traffic pattern flight missions, combined with multidimensional task assessments, to obtain flight performance scores. Selected HRV features, Min_HR (minimum heart rate), SDNN (standard deviation of normal-to-normal intervals), SD2 (long-term variability index in Poincare Plot), Modified_csi (modified cardiac sympathetic index), were identified and used to train classifiers (RF, KNN, GBDT, XGBoost) for pilot mental workload level classification. The XGBoost model demonstrated optimal performance after feature selection, with accuracy increasing from 50.09% to 66.67% (a 16.58% improvement) and F1-score rising from 37.63% to 58.33% (a 20.70% improvement) compared with all HRV feature. The findings revealed selected HRV suppression during high-workload phases (landing) with the lowest performance scores, whereas HRV recovery and peak performance scores were observed in low-workload phases (cruise). This research establishes a reliable framework for real-time pilot mental workload monitoring and provides predictive insights into cognitive overload risks during critical flight operations.