以測試車調查都市幹道旅行速率之誤差分析

Errors in Using Test Cars for Estimating Average Travel Speeds on Urban Arterials

平均旅行速率為評估都市幹道常用之績效指標，多以測試車（test car）調查方式取得現場資料，但利用此法估計平均旅行速率需要蒐集多少樣本是個問題。美國2000年公路容量手冊指出可能要6到12個樣本，而國內根據交通工程手冊辦理「路段行駛時間與延滯」調查，係採用6次測試車來回的資料。因旅行時間的變異性會影響所需之樣本數，但此變異性又無法正確的估計，本研究利用交通部運輸研究所發展之「公路交通系統模擬模式（Highway Traffic Systems Simulation Model）」，以模擬不同測試車次數在各種車流需求及交通控制策略組合情境下之狀況。結果發現，一般旅行時間變異係數在0.16～0.39範圍內時，若採用6趟次測試車資料，要維持估計誤差在2公里∕小時內之機率僅介於35％～95％間；若要誤差不超過2公里∕小時的機率達95％，則測試車之樣本數須在4～30之範圍。

Average travel speed is a measure of effectiveness commonly used for evaluating the performance of urban arterials. A practical method for estimating average travel speed is to float a test car in the traffic of the arterial being investigated. But it is still not clear how many test-car runs should be conducted to obtain reliable estimates of average travel speed. The Highway Capacity Manual published in 2000 in the United States by the Transportation Research Board indicates that 6 to 12 runs may be needed. Taiwan's Traffic Engineering Manual published by the Ministry of Transportation and Communications (MOTC) suggests that 6 test runs be carried out for estimating average travel time and delay. The travel times of vehicles can vary substantially on an urban arterial. Therefore, it is likely that the aforementioned numbers of test runs may not always lead to reliable estimates. To provide a better insight into this issue, this study uses the Highway Traffic Systems Simulation Model developed by the Institute of Transportation under the MOTC to examine the relationship between the number of test runs and the error in estimating average travel speed. The simulation results reveal that the coefficients of variation of the travel times on Taipei's urban arterials are in the range of 0.16 and 0.39. Under the circumstances, the probability that 6 test runs can provide estimates of average travel speed with an error of less than 2 km/h is 35% to 95%. In addition, 4 to 30 test runs are needed to achieve a 95% probability that estimation errors will not exceed 2 km/h.