Title: 高速公路動態車流模式之估計
The Estimation of the Freeway Dynamic Traffic Flow Model
Authors: 高仁風
Jen-Feng Kao
周幼珍
Dr. Yow-Jen Jou
統計學研究所
Keywords: 廣義動差法;有效動差法;半無母數;隨機微分方程式;偏微分方程式;Generalized Method of Moments, GMM;Efficient Method of Moments, EMM;SemiNonParametric, SNP;Stochastic differential equation, SDE;Ito- Taylor expansion;Partial differential equation, PDE
Issue Date: 1998
Abstract: 傳統分析車流理論中, 將車流行為視為波的傳遞, 而使其行為具有動態性。本文將考慮隨機性, 故引入隨機微分方程式 (stochastic differential equation, SDE) 模式來敘述之。但實 際上, 此模式目前尚無法求解, 僅能以模擬方式了解車流, 也尚未討論其指定模式 是否符合實際車流行為。而 Merton (1973), Vasicek (1997), CIR (1985) 曾經用 此連續時間模式探討。正如分析定性的微分方程式一般, 在既定之漂移函數和 擴散函數下, 將模式離散化後, 用廣義動差法估計參數 (Chan (1992)), Nelson (1990) 探討當抽樣區間趨近於零時的極限行為, Lo (1988) 嘗試用最大概似估計法。 Gallant & Tauchen (1996) 提出有效動差法, (EMM), 來估計參數, 並經理論證明出當動差條件及樣 本數增加時, 其估計式效率性近似最大概似估計量, 且其所用準則 (criterion) 會漸近卡方分佈, 並能經此 檢驗模型之正確性。 本文以高速公路一天之資料, 利用 EMM 法, 估計 SDE 模式之參數, 希望藉此證實工作, 指定數個隨機微分方程式, 求得其中最符合資料之模式參數, 並了解估計 SDE 模式時應注意之處以得到一符合實際狀況的車流模式。
In this thesis we use stochastic differential equation (SDE) to formulate the dynamic bahavior of traffic flow and use real data to estimate the parameters in the model. The exact solution and the likelihood function are unknown except for only a few special cases. This fact makes MLE infeasible. Most authors use generalized method of moments (GMM) to estimate parameters. Gallant and Tauchen (1996) proposed a systematic way to obtain the moment conditions (or estimating function) and proved that the estimators so obtained can be as efficient as MLE. The minimized criterion can be used to test system adequacy. We apply the method and diagnostic tests to the number of vehicles passing 86 km spot on Feb. 16 of Northern Second Freeway data is collected every five minutes.
URI: http://140.113.39.130/cdrfb3/record/nctu/#NT870337013
http://hdl.handle.net/11536/64004
Appears in Collections:Thesis