A Case Study on Multi-Agent Based Simulation of Pedestrian Crowds in Outdoor Events of Hong Kong

In this study, the author made efforts to build a multi-agent based system to simulate the pedestrian crowd’s behaviour utilizing GIS data. Taking the fireworks display at Tsim Tsua Tsui of Hong as the sample case, the event was catalogued into different scenarios, arrival and dispersal. Since individual motion is significant to the simulation, different sets of rules were designed for each agent to behave including pedestrian and crowd manager respectively for different scenarios. And the agents will act intellectually based upon the rules after analysis of collected information and communicate with other agents. The entire region was raster into grid cells at a minimum size of space which is suitable for one pedestrian. Once simulation started from arrival, each pedestrian agent will be generated from randomly chosen subway station in this area, and then select one of the fireworks display sites as its favourite attractor agent. During each time step, the agents will search available routes to the attractors and take a step according to the decision. The pedestrian won’t stop acting until they arrive at the destination cells. While in the dispersal, things will be a little different for the exchanged start points and the attractors. Meanwhile, several crowd manager agents were set moving around within the region and will send instructions to the pedestrians to keep them away from overcrowded district they confirmed. On the other hand, the simulation process could be controlled through the interface over different parameters such as the maximum amount of pedestrian agents and allowable density of MTR station. And as essentials of practical rehearsing before the events, the users could also get control of the simulation by means of manually blocking interested streets or opening buildings to discover affected pedestrians behaviors under what-if situations, thus it will be helpful for the uses to prepare for certain emergency and more effective than conventional sand table game. Through the development, some tactical technical issues were discussed, including the path finding and time step slice problem which may be somewhat different and complicated at real practical level.