A generic and hybrid approach for pedestrian dynamics to couple cellular automata with network flow models

Large pedestrian dynamics scenarios are difficult to simulate by models with high spatial resolution in reasonable time, due to high computational costs. Possible solutions are hybrid approaches, which combine models of different spatial scales. On the macroscopic scale, network flow models are used to model pedestrian dynamics. These models reduce the scenario to a simple network of nodes and edges. They simulate quite quickly, but suffer from low spatial resolution. Mesoscopic and microscopic models simulate pedestrians as discrete and singular individuals. On the mesoscopic scale, the movement space of pedestrians is reduced to a cellular grid, while pedestrians on the microscopic scale move on a continuous space. These models have higher spatial resolution for the cost of more computational time. By the use of hybrid approaches, potentially dangerous regions (e.g. bottlenecks) of the scenario can be calculated by models with a higher resolution, while the remaining parts are simulated by less computationally heavy models. This significantly reduces the overall computational effort. We propose a generic approach for pedestrian dynamics that is capable of coupling almost any network flow model with almost any cellular automaton. We use a transition-zone based approach, which is able to transform pedestrians from one model to the other and vice versa without further knowledge about the coupled models. The coupled models simulate separated regions of the scenario and are only connected by shared transition zones. To execute valid transformation, the transition system only needs information about the scenario geometry, the velocity, cell positions on the mesoscopic scale and, from the macroscopic scale, information about the amount of pedestrians on the edges.