A Spectral Approach Towards Analyzing Air Traffic Network - - PowerPoint PPT Presentation

A Spectral Approach Towards Analyzing Air Traffic Network Disruptions (Li et al., ATM2019)

high delay correlation

LGA delays: 90 min BOS delays: 90 min LGA delays: 170 min BOS delays: 10 min

Two knowledge gaps:

• How do we distinguish

unexpected spatial distribution

• f delays when the magnitude
• f the delays are similar (on a

network scale)?

• What spatial delay distributions

are observed for different types

• f air traffic network

disruptions?

180 min total delay Expected distribution

high delay correlation

180 min total delay Unexpected distribution ATL Delays BOS delays ORD delays TPA delays SAN delays ⋯ ⋯ ⋯ ATL Delays BOS delays ORD delays TPA delays SAN delays ⋯ ⋯ ⋯

Our solution: Graph signal processing to analyze spatial delay distributions

Key idea in this paper: Use Graph Signal Processing and total variation of graph signals to characterize expected vs. unexpected spatial delay distributions and relative differences, providing a systematic, low-dimensional, and generalizable methodology.

[Widespread winter storms in ORD, NYC] [Denver snowstorm] [SFO, LAS fog + Augusta Masters]

Important applications and extensions

Airline-specific analysis of spatial delay distribution outliers

Airline 1 Airline 2 …

Phase portraits of the airline recovery process after irregular

• perations

Expected recovery trajectory Anomalous recovery trajectory

Unexpected spatial distribution of delays during recovery (Days 3 & 4) Unexpected spatial distribution of delays during Recovery (Day 4)