本文比較累積機率法與總體停車或率法在設計黃燈需時的異同，經進行不同車輛到達類型下黃燈需時之簡例分析發現，累積機率法適用於一車到達狀況，對多車到達狀況會高估其黃燈需時，而總體停車或率法則可適用於各種車輛到達狀況。復經臺北市一處路口之實地驗證發現，於不同車輛到達類型下，累積機率法的黃燈需時累積機率曲線會出現明顯的差異，而總體停車或率法則相當一致。實地觀察發現黃燈結束後多車到達狀況闖紅燈比例高於一車到達狀況，基於此，本研究建議變換時段宜採固定的黃燈長度與變動的全紅 (視路口寬度、車速及車輛到達狀況加以調整) 來設計。

This paper aims to compare the amber-light durations determined by two approaches: the cumulated stopping probability (CSP) approach proposed by Lin et al (1987, 1988) and the aggregated stopping possibility (ASP) approach proposed by Hsu and Lan (2003). We demonstrated with an example and discovered that the CSP approach worked well under single-vehicle arrival condition, but it overestimated the desired amber duration under multi-vehicle arrival condition; in contrast, the ASP approach worked well for both circumstances. We further conducted a field case study in Taipei. The results show that, under various types of vehicle arrivals, the CSP approach leads to diversified cumulative stopping probability curves of desired amber durations, however the ASP approach comes out with rather consistent results. From the field study, more red light runners (violators) are observed at the end of the amber light cycle under multi-vehicle arrival condition than single-vehicle arrival condition. Therefore, this study proposes a design of fixed length of the amber light and varied length of the red light that can be adjusted according to such factors as intersection width, approaching speed and vehicle arrival types at the light change.