Di ZUO

Education

B.Sc, Statistics, Civil Aviation University of f China, 2012.09-2016.06
M.Sc, Statistics, University of Leeds, 2017.09-2018.12
Ph.D, Electronic Information, BeiHang University, 2021.09-2026.06

Core Research

Research 1: AirTraj-Diff: Generating Air Trajectory with Conditional Diffusion Probabilistic Model.

This research proposes a weather-conditioned diffusion model to learn the probabilistic distribution of aircraft trajectories in terminal airspace by capturing their complex spatiotemporal features. The model generates realistic trajectories under diverse weather conditions, supporting conflict detection, operational performance analysis, and risk assessment.

Main contributions of Research 1:

1) Proposed AirTraj-Diff, a novel conditional diffusion framework for learning aircraft trajectory distributions and generating realistic trajectories.
2) Developed a UNet-based spatiotemporal architecture with a VAE module to extract meteorological features and model weather-dependent trajectory patterns.
3) Validated the framework on real-world datasets, demonstrating its ability to generate high-fidelity trajectories for downstream analysis tasks.

Research 2: An Attention-Based Convolutional Network Framework for Detection and Localization of GNSS Interference Sources.

This research proposes an attention-based machine learning framework for detecting and localizing GNSS interference sources using affected ADS-B data. It combines logistic regression for preliminary regional detection with an Attention Mechanism Convolutional Network (AMCN) for precise localization by capturing aircraft-level features and cross-aircraft spatiotemporal relationships. Experiments on real-world aviation data demonstrate its effectiveness in supporting GNSS interference monitoring and enhancing aviation safety.

Main contributions of Research 2:

1) Developed a multi-index GNSS interference detection method using logistic regression to identify affected ADS-B data and generate reliable subsets for subsequent localization.
2) Proposed an Attention Mechanism Convolutional Network (AMCN) that combines convolutional feature extraction for individual aircraft with attention-based modeling of global relationships among multiple affected aircraft.
3) Reformulated interference-source localization as a relative grid classification task to improve robustness against uncertain and erroneous aviation data, achieving Top-1, Top-3, and Top-5 accuracies of 27.75%, 75%, and 85%, respectively.