Abstract:In response to the increasing risk of urban flooding and waterlogging risks caused by frequent extreme rainfall events, traditional micro-simulation platforms have the limitations such as low computational efficiency and oversimplified behavioral models when simulating the large-scale and high-precision personnel evacuations. The application potential of the GPU-based parallel computing platform of FLAMEGPU (flexible large-scale agent modelling environment for graphics processing units) in addressing this issue is discussed. Firstly, the parallel architecture of FLAMEGPU and its technical advantages in agent-based modeling are analyzed. Subsequently, a refined agent-based model coupling hydrodynamic and pedestrian behaviors is built, which comprehensively considers the behavioral logics such as the individual attributes, the impact of water depth and flow velocity on movement speed, the physical fatigue, the panic emotion, and the probability of instability when wading through water. Finally, the Tongzhou Sub-Center Transportation Hub Station in Beijing is taken as a case study to design and conduct the simulation verification. The results indicate that FLAMEGPU can efficiently complete the computing of intelligent agents at a scale of tens of thousands, and the simulation time is within an acceptable range, which has verified that FLAMEGPU provides an efficient technical path for solving large-scale and high-precision flooding evacuation simulations. Its simulation outcomes can provide the quantitative decision support for optimizing the emergency evacuation schemes for urban infrastructure and enhancing the public safety levels.
