Automatic Perception of Aircraft Taxiing Behavior via Laser Rangefinders and Machine Learning

Understanding aircraft taxiing behavior, including type, speed, and lateral deviation, is essential for optimizing runway design and enhancing airport management. This study introduces a novel monitoring system deploying three laser rangefinders, coupled with computational models and machine-learning algorithms, to evaluate aircraft taxiing behavior automatically. Field tests at Chengdu Tianfu International Airport demonstrated the system’s effectiveness. Identifying aircraft types based on landing gear wheel span can be challenging due to overlapping measurements between certain classes. To address this, a classification model based on convolutional neural network (CNN) model is developed and validated by onboard radar data, which achieved an 80% of accuracy rate in aircraft type identification. Analysis of taxiing speeds during take-off and landing revealed significant variations influenced by runway direction, with longer acceleration or deceleration distances leading to broader speed distributions. Lateral deviation analysis indicated a positive skew in wheel track distributions, suggesting a tendency for aircraft to drift toward one side of the runway. The proposed techniques provide a simple and reliable way for aircraft taxiing behavior monitoring, and the findings offer valuable insights for improving runway design and airport safety.