基于巢式和交叉巢式logit的行人步行行为模型

为量化分析行人行为并仿真行人步行轨迹,建立基于巢式和交叉巢式logit的行人步行行为模型.首先,提出建模思路,重点分析行为理论、NL模型和CNL模型等基本理论.其次,将行人分为自由流和非自由流两种状态,针对保持原方向、距离目的地更近、自由流下的变速、躲避对向行人和超越同向行人5种步行意愿,分别建立效用函数,构建行人步行行为的NL模型和CNL模型.最后,通过视频采集的数据,运用biogeme软件标定模型系数,并与实际的行人轨迹比较分析模型的准确性,发现模型能够以80.26%的准确性拟合实际行人轨迹.结果表明,模型能准确拟合低密度情况下的行人步行轨迹,具有实际应用价值.

A pedestrian walking behavior model based on nested and cross nested logit model

To study on pedestrian behavior and simulate pedestrian walking trajectory quantitatively, a pedestrian behavior model based on nested logit (NL) and cross nested logit (CNL) model is proposed. At first, starting from the modeling approaches, basic theories of behavior theory, NL model and CNL model are analyzed emphatically. Then, in the model, pedestrians are divided into two types of free flow and non-free flow. Different functions are built for the five willingness of keeping direction, toward destination, speed changes, avoiding opposite pedestrian and transcend the pedestrian in same direction, respectively. Furthermore, the NL and CNL model of pedestrian walking behavior is proposed. Finally, the software of biogeme is applied to calibrate the coefficients of the model using the data tracked in the video, and the model results are compared with the real trajectory data to validate the accuracy of the model. Accuracy of the model is 80.26\%. The results show that, the built discrete choice model can simulate the real walking trajectory in the low-density situation with high accuracy.