雷達車輛偵測及衝擊波技術應用於緊鄰路口號誌控制之研究

Abstract

在尖峰時刻，市區或高速公路匝道附近，常常會有數個緊鄰路口的交通擁塞問題，其中號誌控制不當也常是擁堵主因。而在作智慧型的號誌控制中，車輛偵測器的車流偵測能力更是關鍵因子。因此本研究從車輛偵測器開始研究，除了偵測傳統流量、速度之外，並偵測十字路口衝擊波，並應用該衝擊波技術作緊鄰路口的號誌控制。 本研究首先針對雷達車輛偵測器，提出車種、車速的演算法。該演算法以最佳辨識演算法為基礎，結合影像處理來學習，使用支持向量機 (Support Vector Machine)來辨識車種、支持向量迴歸 (Support Vector Regression)來估計車長及分辨車種。以市區道路蒐集到的真實資料驗證，並比較K-mean及線性判別分析法 (Linear Discriminant Analysis)後，証實支持向量機及支持向量迴歸可成功精確地辨識機車、小車、大車及超大車等多種車長及推估其速度。 接著，本研究利用前述雷達偵測器的偵測結果，提出新的三個交通參數:空車、有車及停車，並利用此三參數結合車流理論導出路口衝擊波的偵測方法，而且也在模擬環境成功驗證其可行性及精確程度。 最後本研究，提出一個以傳統觸動控制為基礎的臨界路徑控制方法，該方法以關鍵行車路徑來設計時相、依車流回堵情形動態調整路徑時相最大綠燈時間並在萬一車流在綠燈停止不動時，切換時相以避免路口容量損失。該方法中，並以衝擊波理論為基礎推估各臨界路徑上需求綠燈時間，進而提出最佳化模式，求解各路徑最佳均衡綠燈時間。另外，也在一個實際的緊鄰路口組成的群組路口，模擬運作情形，比起傳統觸動模式有顯著改善。

A complementary metal-oxide semiconductor based radar with sensitivity time control antenna is successfully implemented for advanced traffic signal processing. The collected signals from the radar system are processed with developed optimization algorithms for vehicle-type classification and speed determination. In course of optimization, a video recognition module is further adopted as a supervisor of support vector machine and support vector regression. In the meanwhile, skew training data set and numerous classification scenarios are used to test the classifiers. Finally, the results are analyzed and compared. Beside, this investigation provides two traffic flow detection methods for oversaturated signalized intersection. The first method detects intersection shockwaves by innovative traffic parameters involving stopped duration, moving duration, and empty duration. The second method provides upstream arrival rate and speed by shockwaves, signal timing, and traffic flow model. This research has a contribution to the detection of shockwaves and upstream traffic parameters under over-saturated condition which traditional detectors cannot provide. Finally, a novel actuated critical path control model for designing signal timings on closely spaced intersections is presented in this study. Shockwaves are utilized to dynamically adjust maximal green time for each critical path with unstable traffic demands. Combined with path-based progression, this methodology suggests a novel way to deal with closely spaced intersections. A real network had been exemplified with micro-simulation to illustrate the effectiveness of the proposed method. The numerical example demonstrates a satisfying result compare to ordinary full-actuated scheme.