Traffic Evacuation Simulation Based on Multi-Level Driving Decision Model
Shengcheng Yuan,Soon Ae Chun,Bruno Spinelli,Yi Liu,Hui Zhang,Nabil R. Adam
DOI: https://doi.org/10.1016/j.trc.2017.03.001
2017-01-01
Abstract:Traffic evacuation is a critical task in disaster management. Planning its evacuation in advance requires taking many factors into consideration such as the destination shelter locations and numbers, the number of vehicles to clear, the traffic congestions as well as traffic road configurations. A traffic evacuation simulation tool can provide the emergency managers with the flexibility of exploring various scenarios for identifying more accurate model to plan their evacuation. This paper presents a traffic evacuation simulation system based on integrated multi-level driving-decision models which generate agents' behavior in a unified framework. In this framework, each agent undergoes a Strategic, Cognitive, Tactical and Operational (SCTO) decision process, in order to make a driving decision. An agent's actions are determined by a combination, on each process level, of various existing behavior models widely used in different driving simulation models. A wide spectrum of variability in each agent's decision and driving behaviors, such as in pre-evacuation activities, in choice of route, and in the following or overtaking the car ahead, are represented in the SCTO decision process models to simulate various scenarios. We present the formal model for the agent and the multi-level decision models. A prototype simulation system that reflects the multi-level driving-decision process modeling is developed and implemented. Our SCTO framework is validated by comparing with MATSim tool, and the experimental results of evacuation simulation models are compared with the existing evacuation plan for densely populated Beijing, China in terms of various performance metrics. Our simulation system shows promising results to support emergency managers in designing and evaluating more realistic traffic evacuation plans with multi-level agent's decision models that reflect different levels of individual variability of handling stress situations. The flexible combination of existing behavior and decision models can help generating the best evacuation plan to manage each crisis with unique characteristics, rather than resorting to a fixed evacuation plan. (C) 2017 Elsevier Ltd. All rights reserved.