Prediction of Traffic Congestion Propagation

Modeling and quantifying time-lagged accident-induced congestion using causal inference and bootstrap uncertainty analysis.

2021.01 – 2024.02
NAVERTime-Series Representation Learning
Non-recurrent CongestionPropagation MechanismBootstrap Uncertainty

Motivation

  • The impact of congestion caused by the accident is transmitted to subsequent roads, and this congestion propagation is delayed and manifested on some subsequent roads.
  • Unpredictable delayed event, lack of histofical irregular event data make the pattern of traffic congestion propagation difficult.

Goal

To identify and quantify the propagation mechanisms and time-lag effects of accident-induced non-recurrent traffic congestion.

Methodology

  • Modeling the pattern of traffic congestion propagation caused by non-recurrent traffic accidents.
  • Identify the statistical causal relationship between the accident road and the subsequent road, and use the bootstrap method to quantify uncertainties about lags that delay congestion propagation.