Obstacles in pedestrian simulations
Obstacles in pedestrian simulations
复制标题
行人模拟中的障碍
DOI：
10.3929/ethz-a-004696801
复制
发表时间：
2003
期刊：
影响因子：
--
通讯作者：
P. Stucki
中科院分区：
文献类型：
--
作者：
P. Stucki
这份报告提出了两个新的模型来模拟行人的行为。第一种模型使用Hoogendorn和Bovy发明的势场模型方法，但使用波动算法而不是偏微分方程组来计算场。第二个模型使用环境中的重要点，首先生成可见性图，然后使用Dijkstra的算法生成最小生成树。行人的行走方向和速度取决于三个力，类似于和尔冰的社会力模型。第一种力通过使用势场或计算出的生成树将代理引向其目的地。第二种力量调节代理人之间的相互作用，以确保代理人不会走进其他行人，并试图彼此保持一定的距离。第三种力量保证特工不会走进墙壁，他们试图与墙壁保持一定的距离。两种模型模拟的行人行为非常相似，主要区别在于模型的性能。当使用简单的模拟时，生成树方法比势场模型慢，因为它在创建可见性图上花费了大部分时间。当对不同目的地运行更复杂的模拟时，可以发现生成树模型的优势，因为所有生成树可以基于相同的可见性图，否则需要为每个目的地计算新的势场。通过与Weidmann的实验结果进行比较，验证了模型的正确性。模型保持相当简单，以确保快速有效的模拟。通过苏黎世主车站(700x200m)的大型仿真，验证了这两种模型在大型行人基础设施仿真中的适用性。它进一步展示了行人模拟的一些新方面：适应代理人在特殊地形(楼梯、自动扶梯)上的速度的步行能力图；在多层、多目的地(允许使用“步行到下一个出口”等活动)和评级目的地(允许“购买新票，如果可以在售票处而不是自动售票机买到的话多步行10米”)等活动中的步行模拟。
This report presents two new models to simulate the behavior of pedestrians. The first model uses a potential field model approach invented by Hoogendoorn and Bovy, but calculates the field using a wave algorithm instead of partial differential equations. The second model uses significant points in the environment to generate first a visibility graph and then a minimal spanning tree using the algorithm from Dijkstra. The walking direction and speed of the pedestrians depends on three forces similar to the social force model by Helbing. The first force leads the agent toward its destination by using either the potential field or the calculated spanning tree. A second force regulates the interactions between agents, so ensuring that agents do not walk into other pedestrians and try to keep a certain distance from each other. The third force guarantees that agents do not walk into walls and that they try to keep a certain distance from walls. The behavior of pedestrians simulated with the two models is quite similar, the main difference lies in the models’ performance. The spanning tree approach is slower than the potential field model when simple simulations are used, as it spends most time on the creation of the visibility graph. The advantages of the spanning tree model are found when more sophisticated simulations with different destinations are run, as all spanning trees can be based on the same visibility graph, where else a new potential field needs to be calculated for each destination. The correctness of the models is proved by comparing the average pedestrian speed in a corridor with increased pedestrian density and fraction to empirical results collected by Weidmann. The models are kept rather simple to ensure fast and efficient simulations. Some large simulations of Zurich main station (700 x 200m) are used to show the useability of both models for simulations of large pedestrian infrastructures. It further presents some new aspects for pedestrian simulations: a walkability graph adapting the agents’ speed on special terrain (stairs, escalators), simulations in buildings with several floors, multiple destinations (allowing to use activities like ”walk to the next exit”) and rated destinations (allowing activities like ”buy a new ticket, accept 10m of extra walking if it can be bought at a ticket desk rather than an automata”).
DOI：
--
发表时间：
2003-06
期刊：
IEICE Trans. Inf. Syst.
影响因子：
--
作者：
K. Nishinari;Ansgar Kirchner;A. Namazi;A. Schadschneider
通讯作者：
K. Nishinari;Ansgar Kirchner;A. Namazi;A. Schadschneider