Human-in-the-loop System to Investigate Perception-driven Evacuation Behaviors for Modeling and Simulation: Focused on Radiological Emergency

Abstract

Emergency response plans for radiation accidents in South Korea and Japan focus on the macroscopic scale from the perspective of the national or local governments. However, the Fukushima disaster in 2011 has revealed the existence of gaps between the plan and actual situation. The inappropriate evacuation orders that failed to consider individual behaviors of residents acted as the root cause, and the residents of the indoor sheltering zones voluntarily evacuated, resulting in disorganized mass evacuation. In the case of other disasters such as hurricanes, large-scale evacuation simulations using an agent-based model (ABM) are being developed for planning, but modeling and simulation (M&S) for radiation accidents are constrained by the lack of quantitative data on individual behaviors. To overcome this limitation, this study proposes a method of observing evacuee’s behaviors under the nuclear power plant (NPP) accident by building the virtual reality (VR) based human-in-the-loop (HITL) system. For proof of concept (PoC) about the HITL system, we conduct the test constructed with reference to the Fukushima disaster. Finally, the PoC test results show that the HITL system is suitable for observing evacuation behaviors of human. The quantified data collected from the HITL system can be applied to M&S for an effective emergency management tool.