Combining Avoidance and Imitation to Improve Multi-agent Pedestrian Simulation

摘要

Simulation of pedestrian and crowd dynamics is a consolidated application of agent-based models but it still presents room for improvement. Wayfinding, for instance, is a fundamental task for the application of such models on complex environments, but it still requires both empirical evidences as well as models better reflecting them. In this paper, a novel model for the simulation of pedestrian wayfinding is discussed: it is aimed at providing general mechanisms that can be calibrated to reproduce specific empirical evidences like a proxemic tendency to avoid congestion, but also an imitation mechanism to stimulate the exploitation of longer but less congested paths explored by emerging leaders. A demonstration of the simulated dynamics on a large scale scenario will be illustrated in the paper and the achieved results will show the achieved improvements compared to a basic floor field Cellular Automata model.