The problem of efficient pedestrian simulation, when large-scale environment is considered, poses a great challenge. When the simulation model size exceeds the capabilities of a single computing node or the results are expected quickly, the simulation algorithm has to use many cores and nodes. The problem considered in the presented work is the task of splitting the data-intensive computations with a common data structure into separate computational domains, while preserving the crucial features of the simulation model. We propose a new model created on the basis of some popular pedestrian models, which can be applied in parallel processing. We describe its implementation in a highly scalable simulation framework. Additionally, the preliminary results are presented and outcomes are discussed.
展开▼
