基于人机交互的数据驱动式人群行为建模与仿真研究

项目名称： 基于人机交互的数据驱动式人群行为建模与仿真研究

项目编号： No.61502370

项目类型： 青年科学基金项目

立项/批准年度： 2016

项目学科： 计算机科学学科

项目作者： 罗林波

作者单位： 西安电子科技大学

项目金额： 21万元

中文摘要： 人群仿真是预测和分析复杂人群行为的重要手段。为确保仿真的可靠性，如何有效地获取人群行为相关数据并对人群行为进行建模及验证是人群仿真研究的难点问题。本项目以人群中个体的决策行为作为主要研究对象，以人群疏散场景为研究背景，借鉴人本计算的思想，提出基于人机交互的数据驱动式人群行为建模与仿真方法。首先，研究交互式虚拟场景生成方法，通过用户与虚拟场景的人机交互，获取个体决策行为的观察数据；其次，在数据采集的基础上，提出基于进化算法的个体决策规则识别方法，建立个体决策行为模型；最后，通过建立多智能体人群仿真系统，对仿真结果进行验证，并针对人群疏散场景进行预测分析。本项目将为揭示典型人群场景中个体决策行为的潜在规律提供有效的研究方法，并对公共安全管理具有重要的应用价值。

中文关键词： 人群仿真；人机交互；虚拟场景；数据驱动建模；进化算法

英文摘要： Crowd simulation is an important means to predict and analyze complex crowd behaviors. To ensure the reliability of simulation, how to collect crowd behavior data and make use of the collected data to build and validate crowd behavior model has become a challenging problem in crowd simulation. In the project, we focus on the decision making behavior of individuals in the context of crowd evacuation. By leveraging on the idea of human computation, we aim to build a data-driven crowd behavior modeling and simulation framework based on human-computer interaction. First, we will design interactive scenario generation method and ask users to interact with the generated scenarios for data collection. Second, a decision rule identification method will be proposed using evolutionary algorithm. The identified decision rules will be used to drive the decision behavior model of individuals in crowd. Last, a multi-agent crowd simulation system will be developed and validated. The system will be used for “what-if” study in crowd evacuation analysis. The research work in this project will provide a novel framework to reveal the underlying patterns of human decision behaviors in typical crowd scenarios and it can be potentially applied for safety management applications.

英文关键词： Crowd Simulation;Human-Computer Interaction;Virtual Scenario;Data-driven Modeling;Evolutionary Algorithm