郊區路段微觀混合車流特性研析與模式建立

Abstract

台灣地區道路交通車流量大且機車比例相當高，國外文獻較著重汽車車流模式建立，然汽機車之駕駛型態、跟車行為、車輛操控、與車輛推進等特性均不盡相同。國內對混合車流行為雖已有相當之研究成果，但對郊區公路微觀混合車流之研究仍十分有限。故本研究基於跟車理論、微觀車流理論、運動學原理、模糊理論、行為門檻等理論基礎與文獻評析法、攝影調查法、統計分析法、類神經網路、適應性類神經模糊推論系統及模式參數校估法等研究方法，構建郊區路段微觀混合車流模式。跟車模式係利用攝影調查法蒐集混合車道車輛之跟車行為，依車種與機車區位劃分型態，初步研析郊區公路車流特性、道路幾何特性，以及駕駛行為等，並以適應性類神經模糊推論系統構建郊區微觀混合車流跟車模式。變換車道模式則依車種及與考量目標車道之狀況區分型態，進而分別構建混合車流變換車道之行為準則與偏向角模式。本研究所構建國內郊區混合車流特性之交通車流模式，其研究成果將可提供予智慧型運輸系統，作為郊區路段混合車流號誌控制與車流管理之應用。

In Taiwan, the motorcycle traffic flow volume is high. Most reviewed literature emphasizes automobile traffic flow only. However, the traffic behavior of mixed automobile and motorcycle flow is different. This study is based on car-following theory, microscopic traffic flow theory, kinematic theorems, fuzzy theory, and behavioral threshold model. Statistic analysis, neural networks, adaptive neuro-fuzzy inference systems was employed to analyze rural mixed flow. The data of car-following on rural areas was collected by the video camera recorder. Car-following models were built to simulate various flow conditions. Flow characteristic of mixed traffic, road geometry and drivers’ behavior in rural areas was analyzed. Separate models with adaptive neuro-fuzzy inference systems were constructed. Lane-changing behavior models were also calibrated to reflect different flow combinations. The results of this study can be applied for ITS-oriented signal control and traffic management.