基於車流量之路口交通號誌燈分析 - 以台北市為例

Traffic Lights Analysis Based on Traffic Flow: A Case Study of Taipei City

王麒琨(CHI-KUN WANG)

指導教授 ： 朱鴻棋

朝陽科技大學/資訊學院/資訊與通訊系/碩士(2018年)

摘要

智慧運輸系統是智慧城市中不可或缺的系統之一。智慧運輸系統更是影響城市是否能夠運作順利的關鍵，良好的智慧運輸系統能夠提升整個城市的效能。許多國家皆以投入大量的資源在開發智慧運輸系統，然而台灣對於智慧運輸系統還處於初步階段，對於如何應用設備所回傳的資料並沒有一套可行的機制存在。
本論文提出了一種以系統化的方式來進行交通策略的評估，能夠根據車輛偵測器所蒐集的資料進行分析與判斷。透過基於車流資訊的分析來了解路口、路段之間的關係，根據該關係能夠尋找出更佳的交通號誌燈週期配置，以降低交通壅塞的問題。本文使用群集分析做為資料前處理的方法，群集分析能夠有效依據資料之間的關係將資料分群，提升資料的品質及有效降低分析的時間成本。根據提出的演算法將台北市劃分成多個小區域，並且分析區域內的車流旅行模式，將分析的結果作為交通號誌燈週期配置的考量。最終的結果顯示，經過更加完善的考量，可以減少路口的延滯率，降低交通壅塞的程度。並且透過深度神經網路做為分類模型，能夠有效用簡單的資料進行路口壅塞等級的分類，無須再進行繁瑣的運算。

關鍵字

智慧交通系統 ； 群集分析 ； 深度學習

並列摘要

Intelligent transport system is one of the indispensable systems in smart cities. Intelligent transport system is the key to a smooth operation of the city. Many countries have invested a lot of resources in developing intelligent transportation systems. However, Taiwan’s intelligent transport system is still in the preliminary stage. There is no set of feasible mechanisms for how to apply the information returned by the device.
This paper proposes a systematic approach to traffic assessment that can be analyzed and judged based on the data collection by the vehicle detector. Base on traffic flow, the relationship between road junctions can be analyzed. Utilizing these analysis results, it is possible to find the better cycle of traffic lights. This paper uses cluster analysis as method of data preprocessing. Cluster analysis can effectively group data based on the relationship between data that can improve the quality of data, and effectively reduce the time consuming of analysis. This paper will divide the city of Taipei into multiple small areas according to the collected data, and analyze the traffic flow patterns of these areas. The result of the analysis will as a consideration for the configuration of traffic lights. The experiment result shows that after more comprehensive considerations, it is possible to reduce the vehicle delay rate at the intersection and reduce the degree of traffic congestion. In addition, through deep learning as a classification model, it is possible to effectively classify intersection congestion levels with simple data without the need for tedious operations.

並列關鍵字

Intelligent transport system ； cluster analysis ； deep learning

參考文獻

[1] “ISO/TC 204.” [Online]. Available: https://www.iso.org/committee/54706.html. Accessed on: Apr. 22, 2018.

[2] 交通部智慧運輸系統發展建設計畫, 智慧運輸系統發展建設計畫(106年~109年), Dec. 2016. http://www.its-taiwan.org.tw/upload/file/1703271637520366.pdf

[3] A. Festag, “Cooperative intelligent transport systems standards in Europe,” IEEE Communications Magazine, Vol. 52, Issue 12, pp. 166-172, Dec. 2014.

[4] Y. Lv, Y. Duan, W. Kang, Z. Li, F.Y. Wang, “Traffic Flow Prediction with Big Data: A Deep Learning Approach,” IEEE Transactions on Intelligent Transportation Systems, Vol. 16, Issue 2, pp. 865-873, Apr. 2015.

[5] W. Huang, G. Song, H. Hong, K. Xie, “Deep Architecture for Traffic Flow Prediction: Deep Belief Networks with Multitask Learning,” IEEE Transactions on Intelligent Transportation Systems, Vol. 15, Issue 5, pp. 2191-2201, Oct. 2014.