考量不確定性之路網號誌最佳化問題

Abstract

傳統的號誌設計方法多以單一需求量作為設計依據，實際上交通需求會因各種時空因素而有多種可能，若忽略需求的波動，可能會導致號誌控制的效果不如預期，如高於平均之需求量可能無法於一次綠燈時間紓解完畢。雖然文獻中以獨立路口設計考量需求不確定性的號誌時制，但概念尚未推廣至連鎖號誌，無法一次性考量多個路口。基於以上原因，本研究欲以路網為研究範圍，構建考量需求不確定性之號誌最佳化模型，產生足以面對需求波動之號誌時制計畫，提升路網績效。首先以格位傳遞模式構建車流模型，並提出號誌控制模型，分別建立確定性及強韌性模型；接著以車輛的隨機抵達產生150個需求情境，再透過K-Medoids演算法縮減情境數量，供強韌性模型使用。最終結果顯示，本研究所構建之強韌性模型具有效性，其所產生之號誌時制計畫相較於確定性模型可使路網之總延滯值更小，在三個案例中分別使75.33% 、82.67%及60.0%之情境有較少之路網總延滯，足以顯現其強韌性。

Traditional methods of traffic signal design often use a single value of demand as the basis for design. However, in reality, traffic demand can vary due to various spatiotemporal factors. Ignoring demand fluctuations may lead to suboptimal signal control effectiveness. For example, higher-than-average demand might not be discharged within one green light cycle. Although literature has considered signal timing plan under demand uncertainty for isolated intersections, the concept has not been extended to coordinated signals, unable to simultaneously consider multiple intersections. Therefore, this study aims to develop an optimization model for signal timing plan considering demand uncertainty across a network, producing signal timing plans capable of handling demand fluctuations and improving network performance. First, a traffic flow model is constructed using the cell transmission model (CTM), and a signal control model is proposed, establishing deterministic and robust optimization models, respectively. Next, 150 demand scenarios are generated based on the stochastic arrival of vehicles, and the number of scenarios is reduced using the K-Medoids algorithm for the robust model. The final results show that the robust model constructed in this study is effective. Compared to the deterministic model, the signal timing plans produced by the robust model result in lower total network delay, reducing total network delay in 75.33%, 82.67%, and 60.0% of scenarios in three cases, respectively, demonstrating its robustness.