即時應變事故車流擁塞溢流號誌控制系統雛型研究

Abstract

發展事故車流擁塞溢流（Queue Overflows）交通控制及管理系統係屬推動ITS先進事故管理系統中一項重要課題，針對事故車流擁塞溢流管理系統之建構，除須即時偵測事故產生之擁塞溢流情形外，為了避免事故造成更嚴重交通擁塞，本研究以發展一個可以即時應變事故車流擁塞溢流之號誌控制系統為目標，並期能與相關管理策略配合執行，提高道路的使用效率及機動性。 本研究利用Sheu所建構之動態隨機系統模式為基礎，當事故發生於相鄰兩路口間鄰近路段且車流擁塞溢流超過上游路口上游偵測器情境下，藉由上、下游兩路口即時反應事故車流狀態最適號誌控制策略，進而改善事故車流擁塞溢流情形，本研究構建之動態隨機模式包括(1)狀態方程式、(2)量測方程式、(3)延滯方程式、(4)限制式等四個方程式，並加入號誌控制的相關控制變數且將控制變數視為系統狀態變數的函數，構建一具有回饋特性、可以即時應變事故擁塞溢流之交通號誌控制模式，利用擴展卡門濾波理論（EKF）和線性、二次、高斯理論（LQG）為基礎求解動態隨機控制模式，以能確實掌握交通車流資訊的動態性和隨機性，並彌補現有號誌控制系統無法即時應變擁塞溢流情形的缺失，使事故管理系統更為完善。 在事故車流資料收集方面，因為事故發生時的真實車流資料可能無法完整取得，本研究自行建構「即時反應事故車流號誌控制模擬系統」，模擬事故車流於號誌控制狀況下之車流情況，以產生本研究中所需車流資料。經由統計檢定方法對於模式輸出的相關交通參數估計值進行驗證，並藉由相關號誌控制績效指標驗證，結果顯示本研究即時應變事故擁塞溢流號制控制系統能有效改善事故車流擁塞溢流衝擊，具有良好表現績效。

Real-time incident Queue overflows responsive traffic control and management is a vital issue to the development of advanced incident management systems in ITS . In order to develope an incident queue overflows management system , must either apply real-time queue overflows prediction and develop an incident queue overflows responsive traffic singal control system , to promote traffic more efficinet and mobility . The system specification presume that when an lane-blocking incident occurred on the zone of a given link between two sequential intersections to induce queue overflows beyond the upstream detectors of upstream intersection , by effectively controling the two sequential intersestions’ signals , to reduce the influence of the incident queue overfolws . In the methodology development , the mainly proposed method and related concepts are extended from Sheu’s early reserches . Following specification of system states , a stochastic model which comprises four types of equations , including : 1) recursive equations , 2) measurement equations , 3) incident-induced delay equations and 4) boundary constraints , is formulated . The technologies involved in the algorithm primarily include extended Kalman filter and L.Q.G to resulting the proposed stochastic model . Based on the proposed stochastic model , a stochastic optimal control algorithm is then developed in an effort to update the time-varying control variables as well as lane traffic state variables in the real time . To generate efficiently traffic data used the model tests , an “ real-time incident responsive traffic simulator signal control system “ have developed by this reserch . The preliminary test results indicate that the proposed method is better than other comparisive signal control methods and is promising to accomplish the goal of real-time incident queue overflows responsive traffic signal control .