一個基於細胞網路資料的具方向性之平面交通定位演算法

Abstract

近年來智慧型運輸系統(Intelligent Transportation System, ITS)愈來愈普及化，其重要的一環為即時交通資訊服務系統，是ITS技術中民眾最常接觸與運用的範疇，具有關鍵性的地位，而此一交通資訊服務的基礎在於即時交通資訊之收集。透過細胞網路資料的收集與分析，可以用來估計和紀錄即時的交通資訊。此方面研究主要是針對高速公路的交通資訊，因為高速公路為單向性的封閉式系統，所以分析開放性系統之平面道路的交通資訊為一重要的議題。本研究提出「一個基於細胞網路資料的具方向性之平面交通定位演算法」，收集行動設備之細胞網路訊號，並紀錄其交遞的細胞網路編號和時間點，再運用資料探勘方法分析行動設備所在路段，以進行後續的交通資訊之估計與預測。實驗主要考慮三個不同因子的組合測試：細胞識別碼、細胞連結順序、細胞停留時間長度，結果證明路段分類同時考慮三個因子績效最佳，相較於使用舊有方法只考慮單一細胞識別碼因子，績效提升13%。由此可知，加入細胞連結順序與細胞停留時間長度因子有助於路段判別，因此本論文所提出的具方向性定位演算法可適用於平面交通。

In recent years, the development of intelligent transportation systems (ITS) has become more and more popular. Real-time traffic information services of ITS draw much attention. These services are offered based on the collection of traffic information. Traffic information can be estimated and collected by collecting and analyzing the historical cellular network data. Most studies focused on traffic information estimation of highway. However, highway only has one direction, whereas local road has multiple directions. The study of traffic estimation of local road is an interesting research issue. In this thesis, a directed positioning algorithm based on cellular network data for local traffic is proposed to collect the cellular network signals and record the cell identities (IDs) along with the timestamps of these signals. Data mining technique is used to analyze these IDs and timestamps, and subsequently determine the road segments on which vehicles are. In experiments, three different factors are considered: connected cell IDs, cell order, and cell residence time. The results show that the proposed algorithm has the best performance whenever all three factors are considered. With all three factors considered, the effectiveness of the proposed algorithm is improved by 13% compared to previous studies.