模糊邏輯匝道儀控模式-細胞自動機之模擬分析

Abstract

匝道儀控是高速公路交通控制中最常見且有效的控制策略之一，以控制匝道併入車流量的方式達到維持主線服務水準、減少匝道等候車輛與預防事故發生的效果。因此，在過去三十年來已有許多學者針對匝道儀控進行許多的研究與應用。目前匝道儀控策略主要可分為三類：定時儀控、交通感應儀控(獨立型與整合型)以及可接受車距控制，其中交通感應儀控可適應性的即時反應交通狀況，因而受到許多學者大量的關注。過去已有許多交通感應儀控演算法，例如：ALINEA、SWARM、METALINE、連鎖匝道演算法、線性規劃等，其中大部分的方法都是以即時交通資訊透過數學模式進行儀控率推估。 然而，由於快速且明顯的交通狀況變化，並不適合以明確性方式進行匝道儀控。模糊邏輯控制(Fuzzy Logic Control，FLC)是由if-then規則所組成的專家系統，可以處理含糊不清或不明確的人類認知和判斷，是明確性專家系統所無法處理。因此，本研究首先建構車流型態判斷模式，以進行四種車流型態判斷，包括自由流、輕微同步流、強烈同步流和大範圍擁擠流。以模糊邏輯控制在考慮主線車流型態和匝道等候車輛數下，分別建構獨立型和整合型儀控策略，獨立型儀控策略是在只考慮局部的交通資訊下進行儀控率推估；整合型儀控策略則是加入上游匝道儀控率為狀態變數之一。為了進一步研究且比較不同匝道儀控策略的績效和車流型態的轉變，因此本研究將以細胞自動機(Cellular Automata，CA)發展微觀車流模擬模式。 透過簡例和實例証明本研究所研擬之獨立型和整合型儀控策略之績效，以不實施匝道儀控、定時匝道儀控、ANCONA在不同幾何設計路段和交通情境下進行績效評比。從簡例和實例模擬結果來看，以不實施匝道儀控下的總旅行時間為比較基礎，在簡例中整合型儀控策略的改善程度為2.16~6.66%；在實例中改善程度為7.96%為最佳，其次為獨立型儀控策略。此外，更進一步觀察車輛的時空速率變化，可以發現在實施本研究所研擬之儀控策略下車輛速率的波動較平緩且平均速率較高。由此可知，本研究研擬之儀控策略的實用性和控制績效皆可獲得確認。

Ramp metering is one of the most popular and effective strategy for freeway traffic control. It aims to control on-ramp traffic so as to enhance mainline level of service, reduce on-ramp queue and prevent accidents. Numerous related researches have been conducted and even been field tested for over thirty years. The ramp metering algorithms can be divided into three main categories: pre-timed ramp metering, traffic responsive metering (isolated and integrated), and gap-acceptance merge control. Since traffic responsive metering can adaptively respond to real time traffic conditions, it has received intensive attentions from researchers. Many traffic responsive metering algorithms have been developed, such as ALINEA, SWARM, METALINE, linked-ramp algorithm, linear programming. Most of them employ mathematic models to determine the optimal metering rates by considering real-time traffic information. However, due to the rapid and remarkable fluctuation of traffic conditions, it might be rather risky to control the on-ramp traffic based upon a clear-cut (crisp) judgment and control. Fuzzy logic controller (FLC), an expert system based on if-then fuzzy rules, has the advantages of treating ambiguous or vague aspects of human perception and judgment, with which a non-fuzzy expert system normally cannot deal. Thus, this study first develops a traffic phase determination model to indicate the traffic condition from four phases: free-flow, light synchronized, heavy synchronized, and wide moving jam. Fuzzy logic ramp metering models by considering mainline traffic phase and on-ramp queue length are then developed under two metering strategies: isolated and integrated. The former strategy is to determine the metering rate based the local traffic information alone, while the latter strategy further considers the upstream metering rate as an extra state variable. In order to further investigate and compare the performances and traffic phase transitions of various ramp metering strategies, a microscopic traffic simulation model, cellular automata (CA), is then developed. To demonstrate the performances of the proposed ramp metering models: isolated and integrated, case studies on an exemplified example and a field example of are conducted, respectively. Comparisons with non-metering, pre-time metering and ANCONA metering models under various geometric networks and traffic scenarios are also made. The results on both exemplified example and field example consistently show that the integrated fuzzy logic ramp metering model performs best, which can curtail 2.16~6.66% and 7.96% of total travel time of non-metering model under exemplified and field examples, respectively, followed by the isolated fuzzy logic ramp metering model. In addition, from the in-depth investigation of the temporal and spatial variations of vehicular speed, it indicates that the average speed can be largely increased while speed variations can be reduced under the proposed metering models. Thus, the applicability and performance of the proposed models have been validated.