On modeling and evolutionary optimization of nonlinearly coupled pedestrian interactions

Abstract

Social force based modeling of pedestrians is an advanced microscopic approach for simulating the dynamics of pedestrian motion. The developments presented in this paper extend the widespread social force model to include improved velocity-dependent interaction forces. This modeling considers interactions of pedestrians with both static and dynamic obstacles, which can be also be effectively used to model pedestrian-vehicle interactions. The superiority of the proposed model is shown by comparing it with existing ones considering several thought experiments. Moreover, we apply an evolutionary algorithm to solve the model calibration problem, considering two real-world instances. The objective function for this problem comes from a set of highly nonlinear coupled differential equations. An interesting feature that came out is that the solutions are multi-modal. This makes this problem an excellent example for evolutionary algorithms and other such population based heuristics algorithms. © 2010 Springer-Verlag Berlin Heidelberg.