Resilient transportation network from natural hazards - Development of a disaster prepared roadway planning framework by geographic machine learning traffic simulation Grant uri icon

abstract

  • Due to global warming, coastal and inland areas are being threatened consistently with flash flood or sea level increasing caused by hurricanes or heavy rains that disrupt current road network. In order to guarantee an effective evacuation process, a new roadway plan needs to be supported systematically by machine learning optimal evacuation plan and traffic analysis to prepare upcoming disaster situations.
    There have been a number of research efforts in evacuation planning such as flooded area prediction by geographic analysis, optimal routing to minimize evacuation traffic time, and traffic analysis by large scale simulation. However, an insightful decision-making framework to solve this kind of complex problem in a systematical way has not been reported. For example, planning a new road by considering several disaster related network disruption situations could change the traffic network equivalence in a specific way and eventually make an evacuation process not only efficient but also effective. We propose a framework that incorporates a geographic flood prediction to identify flooded/closed roads, an optimal evacuation routing based on the disrupted road network, and a large number traffic simulation to validate the overall network efficiency. The proposed work will support finding an optimal roadway that works for both normal road traffic as well as natural disaster scenarios.
    The research tasks include, 1) Geographic analysis: gathering data (past flood/land slide data, regional residential vehicle dada), modeling dynamic flood scenarios, identifying probability of road closing events.; 2) Machine learning: dynamic optimal routing based on road network disruption based on probability; 3) Traffic simulation: traffic analysis (importing road network information, modeling evacuation demand and supply, adding new roadways). Please see an overview of the framework in Figure 1.
    The products of the research will be 1) decision supporting framework that include ArcGIS based geography model, 2) dynamic machine learning model, and 3) traffic simulation models that identifies bottleneck areas and provides new candidate roads to prepare effective evacuation process in the area vulnerable to natural disasters.
    Anticipated benefits include, 1) an effective operational benefit to cost ratio by using the proposed framework when planning a new roadway by considering both the normal and the evacuation transportation network traffic performance; 2) a systematically supported road planning decision process by an integrated models from geography, optimal routing, and traffic simulation.
    This project will be led by Dr. Lee in Engineering (1. ArcGIS and 3. transportation simulation) and Dr. Wu in Computer Science (2. dynamic optimal routing algorithm). The research team has the expertise in transportation network simulation and machine learning.

date/time interval

  • February 2023 - February 2024

contributor