Abstract:Collisions during airport surface operations can create risk of injury to passengers, crew or airport personnel and damage to aircraft and ground equipment. A machine learning model that is able to predict the trajectories of ground objects can help to diminish the occurrences of such collision events. In this paper, we pursue this objective by building a spatial–temporal graph convolutional neural network (STG-CNN) model to predict the movement of objects/vehicles on the airport surface. The methodology adopted in this paper consists of three steps: (1) Raw data processing: leverage Apache Spark to parse a large volume of raw data in Flight Information Exchange Model (FIXM) format streamed from the Surface Movement Event Service (SMES) for the purpose of deriving historical trajectory associated with each object on the ground; (2.1) Graph-based representations of ground object movements: build graph-based representations to characterize the movements of ground objects over time, where graph edges are used capture the spatial relationships of ground objects with each other explicitly; (2.2) Trajectory forecasts of all ground objects: combine STG-CNN with Time-Extrapolator Convolution Neural Network (TXP-CNN) to forecast the future trajectories of all the ground objects as a whole; and (3) Separation distance-based safety assessment: define a probabilistic separation distance-based metric to assess the safety of airport surface movements. The performance of the developed model for trajectory prediction of ground objects is validated at two airports with varying scales: Hartsfield-Jackson Atlanta International Airport and LaGuardia airport, under two different scenarios (peak hour and off-peak hour). Two quantitative performance metrics–Average Displacement Error (ADE) and Final Displacement Error (FDE) are used to compare the prediction performance of the proposed model with an alternative method. The computational results indicate that the developed method has an ADE within the range 7.55,9.33, and it significantly outperforms an alternative approach that combines a STG-CNN with Convolutional Long Short-Term Memory (ConvLSTM) neural network with an ADE of [15.79,16.89] in airport surface movement prediction, thus facilitating more accurate safety assessment during airport surface operations.

m-ABCNet: Multi-Modal Aircraft Motion Behavior Classification Network at Airport Ramps

A computer vision framework using Convolutional Neural Networks for airport-airside surveillance

Vision-Based Aircraft Marshalling Recognition and UAM Control Command Generation

A Searching Space Constrained Partial to Full Registration Approach With Applications in Airport Trolley Deployment Robot

Deep-Learning-Based Parking Area and Collision Risk Area Detection Using AVM in Autonomous Parking Situation

6D Pose Estimation of Aircraft Based on Multi-Model Data

Airplane Type Identification Based on Mask RCNN; An Approach to Reduce Airport Traffic Congestion

Deep4Air: A Novel Deep Learning Framework for Airport Airside Surveillance

TRAFFIC CONTROL RECOGNITION WITH AN ATTENTION MECHANISM USING SPEED-PROFILE AND SATELLITE IMAGERY DATA

MODNet: Moving Object Detection Network with Motion and Appearance for Autonomous Driving

Automating Aircraft Scanning For Inspection With A UAV And Reinforcement Learning Technique

Airplane Type Identification Based on Mask RCNN and Drone Images

Multi-regime analysis for computer vision-based traffic surveillance using a change-point detection algorithm

Airport surface movement prediction and safety assessment with spatial–temporal graph convolutional neural network

Parking Lot Occupancy Detection with Improved MobileNetV3

A study on accident prevention AI system based on estimation of bus passengers’ intentions

RAMP-CNN: A Novel Neural Network for Enhanced Automotive Radar Object Recognition

Robustness Assessment of a Runway Object Classifier for Safe Aircraft Taxiing

Airplane Detection Based on Mask Region Convolution Neural Network

Initialisation of Autonomous Aircraft Visual Inspection Systems via CNN-Based Camera Pose Estimation

Autonomous flight strategy of an unmanned aerial vehicle with multimodal information for autonomous inspection of overhead transmission facilities