Microscopic pedestrian movement model utilizing parallel computations

Abstract

Paper presents high parallel pedestrian microscopic movement model based on the pedestrian local space discretization into cells. Every cell gains a value computed from the gradient field of the pedestrian's target and presence of other pedestrians and obstacles. This evaluation can be done independently for every cell what makes the parallelization very effective. Pedestrian takes a step towards cell with best price. Paper describes the movement model in detail, outlines a construction of gradient fields and brings in the experiments comparing parallel and serial implementation of presented model.