i 動態號誌時制控制模型為雙層規劃(bi-level programming)模型之 應用。模型中包含上層模型—動態號誌時制最佳化模型與下層模型— 動態用路人均衡路徑選擇模型，其中上層模型的目標為使系統總旅 行成本最小，而下層模型則希望用路人在擁有完整的交通資訊下， 依據自身旅運成本最小化觀念，其路徑選擇結果能符合動態用路人 均衡路徑選擇模型之均衡條件，兩者構成Stackelberg 競局。 本研究即根據周鄭義(1999)所構建的動態號誌時制控制模型進 行一系列的求解演算法探討。由於在求解過程中必須求取變分不等 式敏感度分析資訊，因此研究中應用最短距離法進行敏感度分析， 並與廣義反矩陣方法作一比較。而透過最短距離法敏感度分析結果， 重新探討一般在求解網路設計問題時所應用的演算法，並針對動態 路網的特性，重新加以修正與改進。研究中提出四種以敏感度分析 為基礎的求解演算法，包括：SDAP、SDAA、GEDO 與LAA 法， 而經由數例的測試，其中以SDAA 法兼具演算效率與效能，未來可朝實證研究繼續發展。 The dynamic signal timings control (DSTC) model is an application of bilevel programming model, including the upper level, dynamic signal timings optimal model, and the lower level, dynamic user equilibrium route choice model. The DSTC model may be described as a Stackelberg game, among the upper level tries to minimum the total travel cost by allocating the green times and determining link capacities. The lower level, based on the fixed link capacities , searches the shortest travel time route for use, which can be mathematically represented by the dynamic user-optimal conditions. In this research, we consider several heuristic algorithms for the DSTC model which is constructed by Chou (1999). In the iterative processes of algorithms, the minimum distance approach is used to obtain the sensitivity analysis information for the dynamic user equilibrium route choice model. Besides we verify the difference between the minimum distance and generalized inverse approach for the equilibrium network flow. Through the derivative information, we analyze four heuristics sensitivity analysis based algorithms, including : SDAP, SDAA, GEDO, and LAA. Numerical examples are implemented. According to the result, the SDAA is better than other methods.
