Remaining Track Miles Estimation: Evaluating Current Operation and AI Assistance Potential

Abstract

In commercial aviation, accurate estimation of the remaining track miles (RTM) during descent is essential for energy-efficient trajectory management. Currently, pilots often rely on heuristics and experience due to the lack of consistent RTM information, which can result in suboptimal decisions. This study investigates the accuracy of RTM estimations made by commercial pilots through a structured survey involving scenario-based assessments across seven European airports. Results show a consistent underestimation bias, with a root mean square error (RMSE) of 9.69 NM. To quantify the potential of data-driven alternatives, a machine learning model based on gradient boosting was developed using ADS-B surveillance and weather data. The model achieved significantly lower prediction errors, with an RMSE of 5.43 NM, particularly outperforming pilots in early descent segments. Feature importance analysis revealed that spatial and trajectory-related variables were key to accurate predictions. The findings suggest that integrating predictive models into flight management systems or pilot decision support tools could improve descent planning and operational efficiency. This study provides an empirical comparison between human and AI-based RTM estimations, highlighting the potential for machine learning to complement pilot expertise in future air traffic operations.