都會區交通路網整體效能提升之研究

Abstract

由於車輛機動化、都市化、人口增加與人口密度提高的趨勢，導致都會區交通路網的壅塞成為世界性的趨勢，交通網路整體效能的提升是交通管理中心的關鍵性任務，交管中心管理者在考量交通需求、交通狀態、歷史資料、交通瓶頸點與預測資料等全盤狀況之後，做出提升整體交通路網效能的決策。然而以傳統的動態交通指派模型(DTA)方法，要提升整體交通路網效能卻可能面對許多的困難，包含交通需求的變化、路網的複雜度、時空變化的交通瓶頸點、缺乏事件應變機制、路網變更須重新定義模型等等問題。 在本論文中提出了一個以提升整體交通路網效能為目的的交通知識框架，考量了數個可以改善交通路網效能的要素，包含建立協同式的交通資訊產生與分享的架構、減低交通旅次需求、找出時空變化交通樣式(STP)與時空變化交通瓶頸點(STB)、利用系統找出的交通知識與專家提供的知識來改善交通路網的效能。此框架分成了交通資訊收集與產生、異質交通資訊的融合、交通知識的萃取與交通管理的應用等四個階段，在論文中提出了兩種即時交通資訊收集的方式，包含從車輛定位應用服務(LBS)資料中分析出交通資訊，與協同式交通資訊產生與分享架構，所收集到的交通資訊存放在交通資訊資料庫(TIDB)中，做為以時空資料探勘技術找出各種時空交通樣式(STP)與時空交通瓶頸點(STB)的資料源。在本論文中，我們將這些交通資訊與從這些資訊中找出的相關交通知識應用到兩個系統: 旅程時間預測系統與先進交通管理支援決策系統，前者提供用路人最短旅程時間路徑的資訊，後者提供路網管理中心有關於提升交通路網效能的交通指派建議。這兩個應用系統都採取了知識庫系統的技術，推論引擎中的推論法則(Rule)是由領域專家協助設計的知識本體論以及發掘的交通知識(STP/STB)所轉換而來，而推論所需要的事實(Facts)則包含了系統收集到即時交通資訊與事件、預測的交通資訊、路網架構資訊與交通法規限制等等。 本研究的貢獻簡單彙整如下:一、建立一個交通資訊與知識的系統框架，為進一步發展智慧型運輸系統的基礎工程；二、提出數個以時空資料探勘方式找出時空交通樣式與交通瓶頸點的方法；三、提出的動態混合式旅程時間預測方法，其精準度比即時資料旅程時間預測與歷史資料旅程時間預測方法都來的高;四、以知識庫方法實現的交通管理決策系統比傳統的交通指派方法更容易實現在都會區路網上，並且系統可以持續更新知識庫以得到更精準的推論法則。

Global traffic network performance enhancement is one of the critical issues for the urban network administration since traffic congestion has been increasing world-wide as a result of increased motorization, urbanization, population growth and changes in population density, especially in urban network. Network administrators try to enhance the traffic network performance by overall considering all the related issues including current network traffic status, traffic demands, historical traffic patterns, traffic bottlenecks, and predicted traffic status. However, difficulties and issues are raised in traditional Dynamic Traffic Assignment (DTA) models, including traffic demand dynamics, network complexity, spatiotemporal traffic bottlenecks, lack of traffic event consideration, and network topologies evolvement problem. In this dissertation, a traffic knowledge framework is proposed for enhancing the traffic network performance by considering four major factors including decreasing the traffic demand, discovering Spatiotemporal Traffic Patterns (STPs) and identifying the Spatiotemporal Traffic Bottlenecks (STBs), resolving the traffic bottlenecks, and creating a collaborative traffic information generation and sharing framework. There are four phases in the propose framework, which are traffic information generation, heterogeneous traffic information fusion, traffic knowledge extraction and traffic information applications. Two real-time traffic information collection schemes are proposed including traffic information derived from the LBS-based applications and collaborative traffic information generation and sharing framework. All the collected traffic information is kept in the Traffic Information Database (TIDB), from which STPs as well as STBs are mined by spatiotemporal data mining techniques. The discovered traffic knowledge is applied to travel time prediction system and Advanced Traffic Management System (ATMS) decision support system, where the former assists the travelers to select the shortest travel time path, and the latter provides traffic assignment suggestions to network administrators for enhancing traffic network performance. Knowledge based system technique is adopted for these two applications with pre-designed domain knowledge ontologies obtained from the domain experts. The collected as well as predicted traffic information, events, network topologies, constraints are regarded as the facts in the inference engine. The contributions of this work can be summarized as: a traffic information and knowledge framework is built for the ground works of ITS, several hypothesis based spatiotemporal data mining methods are proposed for discovering STP/STB, the proposed travel time prediction system outperforms the real-time predictor as well as the historical predictor, the proposed ATMS decision support system is tractable to apply to real urban network comparing to traditional DTA-based approaches, and the knowledge base is continuously incremental updated to get more precise rules.