Levels of Real-Time Crowd Navigation Behaviour in Urban Environments

Abstract

Continuously increasing interest in real-time crowd simulations motivates researchers from multiple disciplines to investigate more realistic and efficient simulations. Often real-time crowd simulation systems are developed to enhance a user s experience. Therefore, in these applications, most of the computer processing power assigned to the crowd simulation should be given to the agents that will clearly gain a user's attention. This paper investigates methods to improve the computational efficiency of the simulation while not effecting the user s experience. For this reason six dynamically sized areas are defined according to the camera's position and direction. Three types of motion algorithms with different complexities are defined to be used in the different simulation areas. The proposed technique is used to simulate 10000 pedestrians in a 4km2 urban environment and is shown to offer more than 10 times improvement in computational performance.