On the modelling of transportation evacuation: an agent-based discrete-event hybrid-space approach

This paper develops an agent-based discrete-event simulation (AB-DES) modelling framework for transportation evacuation by integrating an event scheduling scheme into an agent-based method. This framework has a unique hybrid simulation space that includes a continuous space and a flexible-structured network. This hybrid space overcomes the cellular space limitation in cell-based evacuation models and provides flexibilities in simulating different evacuation scenarios. Based on this AB-DES framework, we create an evacuation AB-DES model using the Parallel DEVS (Discrete-EVent System specification) formalism. We develop an algorithm to employ the event-scheduling approach to eliminate time-step scheduling used in classic agent-based models. Experimental results show that the AB-DES model is significantly more efficient than a pure ABS model while keeping high model fidelity and the same model capabilities including agent cognitive capability, collision avoidance, and low agent-to-agent communication cost. The agents’ cognitive capability and autonomy property as well as the hybrid simulation space differentiate this AB-DES model from classic pure discrete-event models.