A Task Partition Algorithm Based on Grid and Graph Partition for Distributed Crowd Simulation

Abstract

The task partition algorithm is the key issues in distributed crowd simulation. To overcome low execution performance of existing task partition methods, we proposed a partition algorithm based on grid and graph partition. Firstly, the virtual environment is partitioned by uniform grid, gpu threads are introduced to accelerating the calculation. Later, all pairs of neighbor cells of the grid are connected with an edge. Then a connected graph is constructed, the cells are set as vertexes and the edges connect the cells. Finally, the connected graph is split by k-way partition method for task assignment. The experiments prove that the coarser-grained method can provide higher performance than existing algorithms for different population distribution while keeping low cost. Also the max imbalance rate can be kept very low.