連續路口之適應性基因模糊邏輯號誌控制系統

Abstract

基因模糊邏輯控制（genetic fuzzy logic controller, GFLC）係利用遺傳演算法進行邏輯規則之選擇與隸屬函數之校估，可避免主觀設定之缺失。GFLC應用於獨立路口號誌控制之績效已獲驗證，但尚未有應用於連續路口之號誌連鎖控制績效之研究。基此，本研究嘗試將其擴展至兩個路口之號誌連鎖控制。基於連續路口之號誌連鎖原則，本研究分別建立GFLC同亮、GFLC遞亮、GFLC互亮，以及GFLC獨立（兩組GFLC系統分別獨立控制兩個路口）等四種控制策略。再以四種定時控制策略為比較基準：同亮、遞亮、互亮與獨立，利用窮舉法求解最佳定時時制。 本研究先設定一組兩個小時連續兩路口之車流資料，用以進行模式之驗證與比較。結果顯示，GFLC遞亮之總延滯最低，控制績效最佳；GFLC互亮的總延滯最高，績效最差。另再與定時時制比較方面，在同一控制策略下，除互亮策略外，應用GFLC的適應性號誌控制模式皆優於同類型的定時控制策略。整體而言，遞亮方式無論是以GFLC控制或定時制方式，其績效皆優於其他號誌連鎖方式。GFLC遞亮控制績效為八種控制策略中最佳。 在情境分析方面，本研究以一組雙十字型之連續兩路口，設定其東西向分別為「低」、「中」、「高」流量三種等級，再依流量組合分為流量相同之「高-高」、「中-中」、「低-低」與流量有異之「高-中」、「中-低」、「高-低」六種情境。其中東西向兩路口設定為號誌連鎖方向。情境分析結果，GFLC在東西向流量有差異的情況下較具優勢，在流量相近的情況下也相當接近最佳化定時制號誌控制方式。 在實例應用方面，本研究以台北市「中正路-文林路」交叉路口與「中正路-中山北路五段」交叉路口作為實例應用。研究結果顯示，四種GFLC控制方式績效皆優於實際現況調查時制，可減少總車輛延滯達26％∼56％。GFLC遞亮績效仍優於其他三種應用GFLC的控制策略。

Genetic fuzzy logic controller (GFLC) can overcome the drawbacks of conventional fuzzy logic controller (FLC) which has to subjectively set the logic rules and membership functions. GFLC has been successfully applied to an isolated intersection signal control, but no related study has been found in the coordinated signal control for consecutive intersections. Thus, this study attempts to construct an adaptive genetic fuzzy logic signal controller to coordinated control the signal of consecutive intersections. Based on the principle of signal coordination, four control strategies have been developed and compared in this study, which are simultaneous, progressive, alternative and independent (two independent GFLC systems for two consecutive intersections) GFLC signal control strategies. In addition, in order to investigate the performance of GFLC, four corresponding optimal pre-timed strategies are considered, which are simultaneous, progressive, alternative and independent optimal pre-timed timing plans. A fully enumerative method is employed to determine the optimal timing plan for these four pre-timed strategies. Two-hour traffic flow data in two consecutive intersections are assumed to evaluate and compare the performance of these four signal control strategies. The results show that the progressive GFLC signal control strategy outperforms than any other three strategies in term of total delay. In contrast, alternative GFLC performs worst. Comparing to pre-timed timing plan, corresponding GFLC strategy has produced less total delay with exception of alternative one. In the other word, the progressive GFLC strategy performs best among these eight signal control strategies. In order to investigate the performance of these strategies in varying traffic patterns, a total of six scenarios are designed by considering three levels of traffic flows in eastward direction and westward direction, which are high-high, medium-medium, low-low, high-medium, medium-low, high-low scenarios. Also notice that the east-west direction is assumed to be the signal coordinated direction. The results show that GFLC strategies obviously perform better when east-west flow volumes are largely diverse, but perform similarly to optimal pre-timed timing plans as the traffic flows of these two directions are the same. Finally, a field study at the signalized intersections of Zhong-Zheng/Zhong-Shan North Road and Wen-Lin/Zhong-Shan North Road in Taipei City is conducted. The results show that our four GFLC strategies can cut the total delays by 26% to 56% in comparison with the current timing plan, where the progressive GFLC strategy still outperforms then other three strategies.