輕軌優先號誌績效預測模式之研究

Abstract

本研究旨在探討輕軌優先號誌策略對公路延滯的影響。隨著國內輕軌運輸系統的陸續建設和通車營運，優先號誌的應用可以提升輕軌的運行效率並增加民眾的搭乘意願，但同時也可能對公路車流造成額外的延滯。因此，在輕軌優先號誌策略和公路延滯之間需要找到一個平衡點。 為了達到研究目的，本研究使用SUMO交通模擬軟體模擬路口情境，並收集相關數據來建立和驗證預測模型。建立不同統計分析和機器學習方法之模型並驗證了其準確性和可靠性，以探討不同模式下的預測效果、計算時間和操作難易度，以及優先號誌策略對延滯的影響。 研究結果顯示，當實施優先號誌策略時，同向車流的延滯時間減少，但衝突方向的延滯時間增加。在同向綠燈時比較大和衝突方向車流量較多的情況下，優先號誌策略並不適用，以避免對衝突方向的延滯造成更大的衝擊。綜合考慮各個模式結果的呈現，本研究選擇了運算效率高、預測效果良好且具實用性的對數線性迴歸模式作為預測模型，可供相關單位應用來評估公路延滯情況。

This study aims to investigate the impact of light rail transit signal priority (TSP) strategies on road congestion. With the ongoing construction and operation of domestic light rail transportation systems, the application of TSP can enhance the operational efficiency of light rail and increase public willingness to use it. However, it may also result in additional delays in road traffic. Therefore, it is necessary to find a balance between light rail TSP strategies and road congestion. To achieve the research objectives, this study used the SUMO traffic simulation software to simulate intersection scenarios and collected relevant data to establish and validate predictive models. Various models using statistical analysis and machine learning methods were developed and their accuracy and reliability were verified. The study aimed to explore the predictive performance, computation time, and operational difficulty under different modes and examine the impact of TSP on congestion. The results showed that the implementation of TSP strategies reduced the delay time for prioritized direction but increased the delay time for non-prioritized traffic flows. In situations where there is a higher volume of traffic in the non-prioritized direction or a greater ratio of green phases in the prioritized direction, the application of TSP strategies is not suitable to avoid causing greater delays in the non-prioritized direction. Considering the presentation of results from various models, this study selected a linear regression model with logarithmic transformations due to its high computational efficiency, good predictive performance, and practicality as the predictive model. This model can be applied by relevant authorities to assess road congestion conditions.