Explainable LiDAR 3D Point Cloud Segmentation and Clustering for Detecting Airplane-Generated Wind Turbulence

TL;DR

The paper tackles wake vortex detection from aircraft-generated turbulence by leveraging a 3D LiDAR point-cloud pipeline powered by a Dynamic Graph CNN for semantic segmentation, followed by clustering to pinpoint vortex centers. A novel perturbation-based explainability method is introduced to articulate model decisions to regulators and operators, addressing safety-critical transparency needs. Across real (Vienna) and synthetic datasets, the approach outperforms image-based baselines and single-step segmentation methods, with clustering refining center localization and perturbation analyses validating reliability. The work demonstrates that 3D point-cloud processing offers superior accuracy and interpretability for wake vortex tracking, with potential for real-time deployment at airports and broad applicability to atmospheric turbulence analysis.

Abstract

Wake vortices - strong, coherent air turbulences created by aircraft - pose a significant risk to aviation safety and therefore require accurate and reliable detection methods. In this paper, we present an advanced, explainable machine learning method that utilizes Light Detection and Ranging (LiDAR) data for effective wake vortex detection. Our method leverages a dynamic graph CNN (DGCNN) with semantic segmentation to partition a 3D LiDAR point cloud into meaningful segments. Further refinement is achieved through clustering techniques. A novel feature of our research is the use of a perturbation-based explanation technique, which clarifies the model's decision-making processes for air traffic regulators and controllers, increasing transparency and building trust. Our experimental results, based on measured and simulated LiDAR scans compared against four baseline methods, underscore the effectiveness and reliability of our approach. This combination of semantic segmentation and clustering for real-time wake vortex tracking significantly advances aviation safety measures, ensuring that these are both effective and comprehensible.

Figures (6)

• Figure 1: Top: Colored wake vortex generated by an aircraft. Bottom: Aircrafts maintaining a minimum distance during landing to avoid wake vortices.
• Figure 2: Schematic of LiDAR (L) scan measuring wake vortices of a landing aircraft (flying out of the page) perpendicular to the runway. Starboard (Str) and port (Prt) vortices are visible wartha2022characterizing.
• Figure 3: Large eddy simulation of the roll-up process for an A340 aircraft. Multiple vortices shed from the aircraft wings form a pair of two counter-rotating swirls.
• Figure 4: Overview of Our Approach.
• Figure 5: Example scans for poor predictions. The circles correspond to the labeled wake vortices.