A Real-Time Simulator for the Behavior of Large-Scale Crowd over City-Wide Maps using CUDA

摘要

In this paper, a novel simulator is designed and implemented to run the simulation for large-scale crowds in real time in a city-wide map, which can provide useful insights for the relevant authorities in time. In detail, in this simulator, each individual in the large-scale crowd is denoted as an agent, and each agent's behavior must obey the fundamental physical principles. Since the behavior of each agent in the large-scale crowd should be computed, it requires large computing resources. This simulator utilizes CUDA (a GPUs parallel architecture) to improve the performance. In doing so, the simulation with thousands of agents over a city-wide map can be run in real time by using a single pc or laptop equipped with NVIDIA graphics card. Based on the designed simulator, simulations were conducted to show how to draw useful insights from the simulation.