大數據為基礎之高乘載車道績效評估與預測—以國1五楊高架段為例

Performance assessment and prediction of high-occupancy vehicle lanes based on big data: A case of the Wugu-Yangmei viaduct

潘嘉宜(Chia-Yi Pan)

指導教授 ： 鍾智林

淡江大學/商管學院/運輸管理學系運輸科學碩士班/碩士(2019年)

https://doi.org/10.6846/TKU.2019.00805

摘要

國內之國道道路容量有限，於平日尖峰時段或連續假期發生道路擁擠已是常態。為了提供永續之公路運輸服務以減少溫室氣體排放對環境之影響，降低用路人對於道路的需求以及分散或轉移道路流量，交通部高速公路局針對國道疏運措施進行交通管制手段，包含匝道儀控管制、差別費率管制及高乘載車輛管制。政府為紓解國道1號行經桃園地區之壅塞路段，解決五股至楊梅段之道路服務水準下降，拓寬國道1號五楊高架段並同時劃設第一條高乘載車輛專用車道，提供服務予大眾運輸或具共乘行為之使用者，但目前無相關文獻或研究探討該高架段實施高乘載管制之績效。
鑒於我國為發展大數據之相關技術，積極釋出大量開放資料供各界人士下載以加值應用並賦予數據價值與意義，因此本研究根據研究需求並比較各項資料集之項目後，利用網路爬蟲之技術自高公局交通歷史資料庫抓取2018年1月至2018年9月之VD五分鐘動態資訊、VD靜態資訊及路段靜態資訊等三種資料集，以利於區分高乘載車道與一般車道進行分析。而所獲取之資料存儲格式為半結構化資料格式，不利於直接應用於資料分析，須另行將之轉換成結構化資料集，並利用ETL之技術進行資料萃取、轉置與合併資料集，最後解析車輛偵測器編號所提供之資訊，如該一偵測器之設置位置或所偵測之車道是否為高乘載車道，擴充資料集之欄位並將三種資料進行資料關聯。
本研究藉車道管制熱點分析將車道壅塞路段與時段資料視覺化，五楊高架段之現行車道管制措施於平、假日皆彰顯其管制成效之優勢，HOV車道與一般車道之服務水準皆良好，僅於連續假期時，南北向車道在車輛進入管制區域後、離開管制區域前以及匯入或匯出至機場系統交流道等區間，HOV車道之服務水準雖維持於A至C級，但一般車道之服務水準已落在E級或F級，鑒於連續假期較平日及一般假日易發生壅塞熱點，本研究欲利用深度學習預測五楊高架HOV管制區域於連續假期期間之壅塞路段與時段，故整合2018年連續假期之歷史交通資料與氣象資料，將資料集隨機分為80%的訓練資料集與20%的驗證資料集，建立多層感知器模式對訓練資料集進行學習，訓練學習完成後使之輸出2019年端午連續假期之流量、佔有率與速率，將之利用車道管制熱點分析時可知若無事故或道路維修，則將未有壅塞熱點發生。而為使模式訓練的更加理想，本研究建議得加入是否發生事故或道路維修等變數，同時指向性明確地訓練數個年度的連續假期之車輛偵測器所蒐集之歷史交通資料，以此提升預測水準，應能給予規劃交通政策者或相關交通單位更為具體之預測數據參考，使之調整管制策略使道路流量得以轉移，減少壅塞熱點發生。

關鍵字

高乘載車道 ； 五楊高架 ； ETL ； 車道管理熱點 ； 多層感知器模式 ； 深度學習

並列摘要

As the capacity of National Freeway is limited, it is normal to get traffic congestion during peak hours or long weekend. For providing sustainable transportation services in order to prevent greenhouse effect and achieve the purpose of transferring or dispersing traffic conditions, Traffic controls is initiated by National Freeway Bureau to avoid gridlock, including ramp metering, different tolls and HOV (high occupancy vehicle) lane. The government decided to broaden an overpass between Jhong-Li and Tai-Shan to provide an auxiliary lane for high-occupancy vehicles. With the opening of the 1st HOV lane, the network between Wu-Gu and Yang-Mei will reroute some of its lines to ease traffic and cut the travel time. However, there is no research to explore the performance of this case.
In order to develop technologies for big data, the government has released data to people for downloading and giving these data value and meaning. Therefore, according to the research needs and compare the items of each data set, we choose three data sets and use web crawler to gather these data from history traffic database for 9 months in 2018. After extracting, transforming and loading data sets, we can analyze the information provided by the vehicle detectors’ ID number and increase the fields to correlate these sets of data.
The study is used lane management hot spots to visualize the traffic congestion during the intersections and time slot. These plots show the overpass’s LOS is comfortable to the users except long weekend. Based on this reason, we combine traffic data and weather data for using deep learning technologies which is MLP (multilayer perceptron) to predict the congestion-prone intersections and time on long weekend. After the training is completed, it will output the volume, occupancy rate and speed. Then we can use lane management hot spots to analyze congestion-prone intersections and time slot. The result is the overpass’s LOS is comfortable to passengers on Dragon Boat Festival long weekend. For improving the model and make it train data better, we suggest to improve forecasting by adding accidents variables, road maintenance variables and gathering several years traffic data for MLP to train data. It will be the reference for transportation planners to adjust the traffic control strategy to transfer traffic flow and reduce the occurrence of gridlock.

並列關鍵字

High Occupancy Vehicle Lanes ； Wugu-Yangmei Viaduct ； ETL Data ； Lane Management Hot Spots ； Multilayer Perceptron ； Deep Learning

參考文獻

英文文獻

1. Cassidy, M. J., Daganzo, C. F., Jang, K. & Chung, K. (2006). Empirical reassessment of traffic operations: Freeway bottlenecks and the case for HOV lanes.

2. Cassidy, M. J., Jang, K., & Daganzo, C. F. (2010). The smoothing effect of carpool lanes on freeway bottlenecks. Transportation research part A: policy and practice, 44(2), 65-75.

3. Chen, C, Varaiya, P. & Kwon, J. (2005). An empirical assessment of traffic operations. In : Mahmassani, H.S. editor, Transpn and Traffic Theory, Procs of 16th International Symp on Transpn and Traffic Theory.

4. Chu, C. P. & Chu, C. I. (2016). High-Occupancy Regulation Policy: Case of National Freeway NO. 1 in Northern Taiwan. 都市交通, 31(1), 43-69.