行人流行人跟隨行為特性與模式之研究

Abstract

本研究目的在建構一微觀行人跟隨模式，以探討影響行人跟隨行為之顯著變因，及分析行人跟隨行為之特性。經現場錄影觀察，符合本研究所需之行人跟隨行為樣本占所有觀察行人數目的7%。並由資料發現行人行為與領導者行人速度、跟隨者行人速度、行人速度差、跟隨距離、起始跟隨距離、性別、跟隨時間等影響變數有關，本研究依觀察資料首先引入跟車理論之模式，再建構線性行人跟隨距離模式及線性行人跟隨速度模式，並進行比較。以預測行人於跟隨行為發生後之跟隨距離，並與現有車流理論中跟車模式比較，發現本研究所建構之行人跟隨距離模式，模式解釋能力R2為0.511，平均絕對值誤差率僅0.122，在四種模式中的預測誤差為最小。顯示線性行人跟隨距離模式能有效預測行人跟隨行為。

The main purposes of this study are to construct a microscopic pedestrian-following model and analyze the characteristics of pedestrian-following behavior. The data of pedestrian-following behavior was investigated in commercial area in Taipei. Statistical tests show that the significant factors affecting pedestrian-following behaviors include leader’s speed, follower’s speed, following distance, gender and following time. First of all, study is going to use two models of car-following theory, to see if they could describe pedestrian-following behavior, and to construct new linear pedestrian-following distance model. After comparing the results, we found that the R2 of new pedestrian-following distance model is 0.511, and MAPE is 0.122. The linear pedestrian-following distance model can predict well with the least error over all four models we tested. It is said that pedestrian-following distance model which we constructed can satisfactorily capture the nature of pedestrian-following behavior.