汽機車混合車流模擬—新細胞自動機模型之研發

Abstract

細胞自動機模型可分類為一種微觀的數值模擬車流模型，近年來世界各地學者曾提出多種不同的細胞自動機車流模型，用以模擬現實生活中各種複雜的交通現象。但由於目前細胞自動機模型主要應用相對稀疏之格點系統，導致其無法應用於都市之交通模擬，因此本研究提出一種全新的細胞自動機模型，以克服上述限制，並將細胞自動機模型之應用範圍擴展至都市交通之模擬。 本研究基本上可分為三部份，首先，本研究提出一個具細微格點系統之細胞自動機模型，以提昇進行交通模擬時之模型解析度，並有效反應微觀的交通車流特性。與現有其它細胞自動機車流模型相較，本研究所提出細微格點系統之細胞自動機模型主要有三項明顯差異；首先本研究提出「共有單元」的概念做為描述不同車輛尺寸及不同道路寬度的共同基本單元，而一單獨細胞單位及網格則分別定義為車輛及道路之基本單元。其次本研究提出立體通用交通參數之觀念，如此於進行交通模擬時可將不同車輛尺寸及不同道路寬度之影響納入評估。以上述細微格點系統之細胞自動機模型為基礎，本研究並建議採用線段線性變化之車輛速度變化方式，接下來本研究並將車輛之減速能力亦納入交通模擬之分析範圍，以改正現有多數細胞自動機模型模擬結果中普遍存在的一個重要瑕疵，亦即當車輛遭遇障礙物或抵達塞車車陣之後緣時會有不合理的急速刹車行為。本研究所構思的減速機制事實上源起於1950年代美國學者Pipes及Forbes等人所提出原始車流理論。後續模擬結果證實，本研究所提出之減速機制已成功改正現有細胞自動機模型上述車輛不合理減速現象。 其次，本研究建立於應用上述細微格點系統之細胞自動機模型進行交通模擬時擷取區域路段交通參數數據之有效方法。模擬結果亦證實透過上述方式所擷取之區域路段交通參數資料除可成功印證德國知名學者Kerner於2004年所提出之三相車流理論外，並可描述三種車流車相間之相變化。 在本研究的第三部份中，藉由上述改良之細微格點系統，對由汽車及機車所組成之混合車流形態之交通特性進行模擬及分析。模擬結果證實，本研究所提出之細胞自動機模型可模擬一些重要的都市交通之混合車流現象，如機車於塞車車陣中之橫向移動或由車陣後方兩輛並排車輛間空隙鑽入車陣之複雜行為。 雖然現有之細胞自動機車流模型仍廣受認同，但仍持續有人質疑其應用範圍多限縮於高速公路交通模擬，且無法由微觀角度成功描述細微之車輛行為。反之，透過模擬結果分析，可印證本研究所提出新細胞自動機模型之優越性，因其成功地克服一般現有細胞自動機車流模型之適用限制，將可用以分析實際生活中多變的交通現象。基於本研究所提出新細胞自動機模型具有高解析度及高應用彈性等優點，未來各種不同的交通現象，如由不同種類車輛所組成之混合車流形態及複雜的都市交通現象，無論由巨觀或微觀觀點，皆可藉之進行分析模擬。期待本研究所建立新細胞自動機模型未來成為細胞自動機模型發展歷程中之一項重要突破，並加速交通模擬技術之發展。預期未來各種不同之交通管制策略於實際採用前將可應用本研究所建立新細胞自動機模型於事前評估其效果。

Recently, various cellular automaton (hereinafter referred to as CA) models that can be categorized as one branch from the microscopic viewpoint, were developed to describe the phenomena of real traffic flows characterized with complex dynamic behaviors. However, the coarse cell unit adopted in existing CA models makes them extremely difficult to implement urban traffic simulation. Therefore in this study, a novel CA model is developed, in order to release this restriction and henceforth expands the application of CA model to urban traffic scenarios. This study basically can be divided into three-folds. First, as the onset, a CA model with refined cell/site system is developed in order to enhance the simulation resolution and henceforth to efficiently gauge the microscopic traffic characteristics. Here three important amendments, as compared with traditional CA models, are carried out. First of all, the concept “common unit” is defined to serve as the basic unit for simulating both roads of various widths and vehicles in various sizes, where “cell” and “site” represent the basic unit for vehicle and roadway space respectively. Following that, the 3-D generalized traffic parameters are defined so as to take vehicular and/or roadway widths into consideration. Coupled with this refined cell/site system, piecewise-linear speed variation is introduced into CA simulations. Following this, limited deceleration which in essence arising from Pipes and/or Forbes’ car-following concept, is also proposed for the sake of rectifying one common defect in most existing CA models—unrealistic abrupt deceleration as vehicles encounter stationary obstacles or upstream front of traffic jams. It is evidenced that the proposed refined CA model successfully fixed the unrealistic deceleration behaviors. Next, the methodology for deriving local traffic parameters when implementing our refined CA models is also defined. The simulation results show that through the derived local traffic parameters, the renowned three-phase traffic patterns, as proposed by Kerner (2004), and the phase transitions among them can be successfully simulated. In the third part of this study, based upon the aforementioned refined CA model, mixed traffic comprised by cars and motorcycles are analyzed. The simulations reveal that the proposed CA model successfully gauges some important traffic characteristics of urban traffic, such as the unique transverse drift behaviors of motorcycles when break inside traffic jams and the lateral drift behavior for motorcycles breaking into two moving cars from the upstream front of traffic jam. Aside from their popularity, the existing CA models in the past were continuously criticized for their significantly biased application to freeway traffic and their failure to uncover delicate vehicular behavior from microscopic viewpoint. On the other hand, our simulations evidence the superiority of the proposed novel CA model since it successfully liberates the above-mentioned restriction and is able to capture the violate traffic phenomena in real world. Based upon the enhanced resolution and the increased flexibility of the proposed CA model, analysis of different traffic contexts, such as the mixed traffic comprised by vehicles of various sizes and sophisticated traffic phenomena in urban area, either from microscopic or macroscopic perspective, will be practical in the future. Thus the proposed novel CA model can be deemed as a breakthrough progress in development of CA model and shed some light for the future analysis of traffic modeling. It is looking forward that via the proposed CA model different traffic control strategies for separate traffic contexts can be efficiently evaluated before practical implementation.