B208 セル上に配置されたマルチエージェントによる群集行動分析の一手法(人の流れ)

概要

論文の詳細を見る
In this report, we describe the new crowd behavior analysis method which combines the multi-agent (MA) with the cellular automata (CA). The features of our simulation method are that the behavior of a crowd is formed by autonomous action of heterogeneous agents, and that the action is interrupted by asynchronous intervention from environment. A crowd behavior is a kind of the complex system by which an unexpected behavior emerges. The behavior is caused by interaction between each agent. Behavior of disorder crowd causes big unexpected accidents frequently. In order to cope with this problem, several precedence researches have performed the simulation of the crowd behavior, and have tried to demonstrate its interesting phenomenon. These researches results have been expected to be useful for planning of a safety building design or security management. However, the conventional methods are not enough to simulate complicated real world's behavior. For example, the following two processes were not considered sufficiently by conventional researches, (a) An autonomous action process of heterogeneous agents, (b) An interruption process of a asynchronous intervention from the environment. In order to add these process to a crowd behavior model, we consider the multi-agent system put on the cell. To realize process (a), we introduce agent attribute parameters such as a physique which affect the behavior of each agent. Next, to realize process (b), we divide agent behavior into autonomous behavior and passive one. The autonomous behavior of each agent is controlled by "action determination processing," and "interruption processing using CA" interrupts that behavior asynchronously. In this model, each agent is located on a continuous plane, and its position is represented by coordinate of real number . Based on agent attribute parameters, which are different in every agent, and mutual position relations with other agents at that moment, an agent begins to move on the continuous plane. On the other hand, a cellular plane is an aggregation of cell which the continuous plane was divided into in 1 square meter unit. Each agent belongs to one of cell which corresponds spatially, then each cell has state condition which is determined by the attribute parameters and the number of agents who belong in that cell. When a value calculated from this state satisfies certain conditions, an interruption processing to the agent belonging to the cell occurs. We develop the simulator based on these techniques, and analyze crowd behaviors, in particular, collided crowds ones. By simulation experiments, the following results are obtained. (1) When agents don't have personal variety, "whirlpools" and the "waves" are observed, and walking speed of agents becomes slow. (2) When agents have personal variety, the crowd constructs "platoons," and flows smoothly. (3) Moreover, the behavior changes with the ranges of a agent's view. (4) The interruption processing generated by change of a cell state is effective in reappearance of actual crowd behavior. It thus is concluded that using the simulation system implemented by this method, the interesting behavior of the movement of the actual crowd behavior can be reproduced. Furthermore, because our model has separately an autonomous behavior process and an interference process , it is expected to be useful for combining this model with the building design models such as IFC (Industry Foundation Classes).
社団法人日本機械学会の論文
2001-11-14

B208 セル上に配置されたマルチエージェントによる群集行動分析の一手法(人の流れ)

スポンサーリンク

概要

著者

関連論文

スポンサーリンク