為了解決道路安全的問題,首要工作是深入了解車流行為,進而提前得知車流行為的動態變化情形,以提供即時安全控制的決策判斷所需資訊;若欲深入瞭解混合車流行為的渾沌現象,則無法忽略機車車流與汽車車流並存於道路上的混合車流情況。本文即探討與分析道路上混合車流行為的時空變化,進而實現車流預測行為。 本文先利用渾沌理論的相空間概念,以描述混合車流環境中各車流型態的行為;再者,結合渾沌理論的相空間概念與人工類神經網路的預測能力,分別構建混合車流參數預測模式、機車車流參數預測模式及汽車車流參數預測模式,並進行分析與預測。為了進行模式驗證,透過計算預測輸出值與實際值之間的均方誤差與相關係數,以評估輸出結果的準確度。 經過比較分析後,除了發現混合車流的渾沌現象較其他車流更為顯著,在預測準確度方面,混合車流參數預測模式較機車車流、汽車車流預測模式準確,即描述車流行為的能力更佳,有助於車流趨勢的預測。
In order to improve road safety, first thing we have to do is to understand the spatiotemporal mixed traffic flow condition for evaluating whether the traffic condition is safe or not. The information is essential for real-time safety control system because it can help real-time safety control system obtain the correct decision. However, mixed traffic flow primarily consists of motorcycle traffic flow and car traffic flow. Therefore, this study will try to discuss and analyze the spatiotemporal changing behavior of mixed flow in detail. First step is to describe mixed flow behavior with chaos theory. Next, the chaos artificial neural network of mixed traffic flow, motorcycle traffic flow and car traffic flow are constructed with neural network theory. In order to ascertain if the prediction ability of the chaos artificial neural network is accurate or not, the mean square error and correlation coefficient are adopted for being the indicator of evaluation. Through the comparative analysis, it is found that chaos phenomenon of mixed traffic flow is more obvious than that of motorcycle traffic flow and car traffic flow. Furthermore, the prediction ability of mixed traffic flow model is more accurate than the other models. It is very essential for forecasting the trend of mixed traffic flow changing condition.