號誌化交叉路口黃燈對駕駛者決策行為之研究

Abstract

號誌化交叉路口幾何設計及其號誌系統規劃的兩大基本目標為安全與 效率，當兩者能同時達到時，便可稱該項設計臻於完美境界；然當安全與 效率無法兼得時，安全便成為交通工程師所考慮的重點。 基於安全是 交通設計最主要考量的因素，本研究將探討號誌化交叉路口於黃燈始亮時 位於疑慮判斷區中之小型車與機車駕駛者的行為，分析駕駛者在行駛過程 中遇上號誌變為黃燈時，其所考慮的因素與採行的決定。由於人的行為相 當複雜且充滿不確定性；因此，模糊理論的應用將是本研究的一個重心， 以便能確實捕捉駕駛者的決策行為。 本研究首先訪查駕駛者對黃燈始 亮時判斷是否通過路口之影響因素，並以多評準群體決策方法求出各影響 因素的權重；繼而利用現場錄影實地觀測，以模糊統計方法建立各因子對 駕駛者影響的隸屬函數，並以適當的模糊運算與模糊推論來結合各因子隸 屬函數，進而構建出駕駛者於駕駛中遇黃燈時的決策行為預測模式。 問卷調查結果發現駕駛者對於黃燈始亮時判斷是否通過路口的影響因素， 其重要度前三名依序為1.行駛車速2.距停止線距離3.駕駛者預期心理。對 各因素分別建立其隸屬函數時，小型車的資料明顯趨近於ｓ型與ｚ型，而 機車的資料則大致呈現上升型或下降型。最後，模式驗證結果顯示，本研 究所構建的駕駛者決策行為預測模式，對於小型車駕駛者而言有較佳的判 中率(達85.6％)，而機車駕駛者的預測結果則尚可接受 (為67％) 。歸結 造成推論誤差因素，可能有1.極端值的影響2.駕駛者預期心理3.迴歸檢定 不完全為線性關係4.機車樣本變異較大 5.車輛本身特性差異等等。至於 在模式未來的應用上，將可供政策規劃者與交通工程師對於相關交通設施 與人因工程在運輸系統上之應用。 The two objections of intersection design and its signal are safety and efficiency. When both of them have come true at one time,we can say the designs are perfect.However,if they can't obtain at the same time, safety will be the most important thing for engineers. For safety is the major factor traffic engineering design considered for traffic engineers, so this study will research drivers' behavior, when the amber light onset, being dilemma zone at signalized intersection. Analyzing what factors are considered and what actions are taken for drivers, when driving on the road, the light change from green to amber. Human behavior is complex and full of uncertainties, so the application of fuzzy theory will be the importance of this study, in order to capture drivers' behavior precisely. First,ask drivers by questionnaire for what factors are important when driving to the cross road on amber light onset, and then use group multiple criterion decision making calculating the weight of each factor. Next by video tapes reviewed, use fuzzy theory to con- struct membership function of every factor, then com- bine with all factors' membership functions,can obtain drivers' decision behavior model. By asking with drivers, speed, distance from stop line and driver' desire mental are three important factors affect driver' decision,amber light onset. The membership functions of car having better shape than motorcycle'. The last consequences show these drivers' behavior models which constructed for car have more better effects than motorcycle' s. It can be concluded the factors affect the precise of models may be 1.ex- treme samples 2.drivers' desire mental 3.data is non- linear 4.motorcycle's samples have much more variation 5.the differences between car and motorcycle, etc. The future application of model will provide human factors and information for engineers in traffic facilities or transportation system using.