基于人工蜂群算法的复杂场景路径规划研究

发布时间：2018-06-16 07:13

本文选题：三维场景建模 + 人群疏散仿真　；参考：《山东师范大学》2017年硕士论文

【摘要】：伴随世界经济、人口的快速发展及城市化进程的稳步加快,随之而来的公共安全事故日益增多,公共安全的管理问题引起社会各界的广泛重视。对人群的疏散行为进行研究,总结行人运动规律,从而在紧急情况下提供直观的疏散路径规划、制定合理的应急预案、保证人民生命财产安全,已成为安全管理中亟待解决的重要问题。在传统演练方法多方面受限的情况下,计算机仿真技术的研究与应用提供了有效的解决思路,克服了传统方法安全性低、效果不真实等问题,能够实现逼真、高效的人群运动行为模拟。对人群的运动行为进行仿真,其内容主要包括三维模型建模、人群运动以及虚拟场景下的宏观路径规划三方面内容。针对以上三方面内容,本文分别提出了相应的人群疏散行为模型及三维模型建模方法。对于传统模型建模方法中存在的成本高昂、操作难度大等问题,采用了三维建模软件进行三维场景模型及虚拟人物模型的构建及贴图工作,该方法能够有效实现不同类型场景、人物的比例还原,同时满足了模型建模真实准确的要求。此外,目前的人群疏散行为模型研究中,还存在宏观路径规划算法适用性差、不支持复杂场景的规模群体运动、行人运动行为不合理等问题。因此,本文提出一种改进的社会力模型与人工蜂群算法相结合的方法,利用改进的人工蜂群算法进行实时宏观路径规划,底层结合修正的社会力模型指导个体运动,应用于不同场景下的人群运动仿真中。本文主要工作及创新点如下:1.对软件构建的三维场景模型进行拓扑结构的求解以及实现真实感虚拟人物模型的不同动作。通过提取三维场景的语义信息获得场景的区域划分关系,从而得到场景模型的拓扑结构,简化场景存储结构和宏观路径规划复杂度,实现复杂场景下路径信息的获取与存储。使用Motion Builder软件对构建的人物模型添加运动,实现虚拟人物在仿真中走路、奔跑等多种行为效果,获得真实生动的虚拟人物模型,为人群运动仿真研究中的动作建模方法做有效参考。2.提出一种引入视觉影响的动态避障社会力模型进行人群的运动行为模拟。在原始社会力模型基础上引入视觉影响因子,体现不同视角对个体运动产生的不同影响,对不可见行人之间的作用力进行限定,使个体运动行为的模拟更加真实合理;对运动过程中存在障碍物的情况,提出在障碍物包围盒顶点设置临时目标点的策略,实现个体在可视范围内的提前避障行为,驱动行人平滑、流畅的完成局部运动过程。3.提出一种基于分组策略的改进引领路径选择机制的人工蜂群算法,实现复杂场景下的实时宏观路径规划。对原蜂群中蜜蜂进行分组,实现只在同组内进行引领选择、信息传递,采用并行计算加快算法收敛速度;改进了引领蜂路径选择机制,兼顾出口拥挤度与路径长度等因素,模拟出疏散中人群尽快逃离的心理现象,提高人群在场景中的实时感知能力,更符合真实的人群疏散效果。结合承担的科研项目,将以上研究成果应用于人群运动仿真系统与真实感渲染平台中,实现不同类型场景下的人群运动模拟。系统及平台实现了场景语义信息提取、宏观路径规划、人群运动及真实感渲染四项功能。通过在系统及平台中进行简单场景和复杂场景两类仿真实验,对疏散仿真效果进行分析与研究,对比真实人群的运动行为特性,能够验证本文所提出的方法能够真实、有效地还原现实场景中的建筑物外形和内部结构,以及逼真地模拟真实人群的运动行为,对于未来研究人群行为特性分析具有较高参考价值和应用价值。
[Abstract]:With the world economy, the rapid development of population and the steady acceleration of urbanization and the increasing number of public safety accidents, the management of public safety has aroused wide attention from all walks of life. The evacuation behavior of the crowd is studied and the rules of pedestrian movement are summarized, thus providing an intuitional evacuation route under the emergency. Making a reasonable contingency plan to ensure the safety of the people's life and property has become an important problem to be solved urgently in the safety management. Under the circumstances of the traditional practice, the research and application of computer simulation technology provides an effective solution to solve the problems of low security and untrue effect of traditional methods. To simulate the movement behavior of the crowd, the movement behavior of the crowd is simulated. The contents of the crowd are mainly composed of three aspects: 3D model modeling, crowd movement and the macro path planning under the virtual scene. According to the above three aspects, the model of crowd evacuation behavior and modeling of three-dimensional model are proposed in this paper. Method. For the high cost and difficult operation of the traditional model modeling method, the 3D modeling software is used to construct the 3D scene model and the virtual character model. This method can effectively realize the different types of scene and the ratio of the characters. At the same time, it satisfies the real and accurate model of model modeling. In addition, in the present study of crowd evacuation behavior model, there are still some problems, such as the poor applicability of the macro path planning algorithm, the scale group movement which does not support the complex scene and the unreasonable pedestrian behavior. Therefore, this paper proposes an improved method of combining the social force model with the artificial bee colony algorithm, and uses the improved artificial bee colony algorithm. The method carries out real time macro path planning, the bottom layer combines the modified social force model to guide the individual movement, and applies it to the crowd motion simulation under different scenes. The main work and innovation points are as follows: 1. to solve the topology structure of the 3D scene model constructed by software and to realize the different movements of the realistic virtual character model. The semantic information of the 3D scene is extracted to obtain the regional partition relation of the scene, thus the topology of the scene model is obtained, the storage structure of the scene and the complexity of the macro path planning are simplified, and the path information is obtained and stored in the complex scene. The motion of the figure model is added to the constructed figure model by using the Motion Builder software to realize the virtual character. Walking, running and other behavior effects in the simulation, a real and vivid virtual character model is obtained, and an effective reference for the action modeling method in the crowd motion simulation study.2. is put forward to simulate the dynamic behavior model of the dynamic obstacle avoidance social force introducing the visual influence to the crowd. The visual shadow is introduced on the basis of the original social force model. The noise factor reflects the different effects of different perspectives on individual movement, limits the force between invisible pedestrians, and makes the simulation of the individual movement more realistic and reasonable; the strategy of setting temporary target points in the vertex of the barrier encircling box is put forward to realize the visible range of the individual in the course of the existence of obstacles in the process of movement. Early obstacle avoidance behavior, driving pedestrians smooth and smooth finishing local movement process.3. proposed an artificial bee colony algorithm based on grouping strategy to improve the path selection mechanism, realizing real time macro path planning under complex scene. By using parallel computing, the convergence speed of the algorithm is speeded up, and the mechanism of leading bee path selection is improved, and the factors such as the exit congestion and the path length are taken into consideration, and the psychological phenomenon that the crowd escapes as soon as possible is simulated to improve the real-time perception ability of the crowd in the scene, which is more in line with the effect of the real crowd evacuation. The research results are applied to the crowd motion simulation system and the realistic rendering platform to realize the crowd motion simulation in different types of scenes. The system and platform implement four functions of scene semantic information extraction, macro path planning, crowd movement and realistic rendering. The simple scene and complex scene two are carried out in the system and platform. The simulation experiment is used to analyze and study the effect of evacuation simulation. Comparing the behavior characteristics of the real crowd, it can verify that the proposed method can truly, effectively restore the shape and internal structure of the building in the real scene, and simulate the real crowd's movement behavior, and study the behavior characteristics of the crowd in the future. The analysis is of high reference value and application value.
【学位授予单位】：山东师范大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.9

【参考文献】