藉由考慮旅次時間的狀態空間模型去估計起迄旅次矩陣

Abstract

在交通管理分析上，起迄旅次矩陣扮演著越來越重要的角色，因此也吸引了很多估計起迄旅次矩陣方面的研究。動態的起迄旅次矩陣估計相當地新。許多動態方法的應用和發展較靜態方法來的廣泛。在這篇論文中，我們提供了方法去估計起迄旅次矩陣。我們使用時間序列中的非高斯狀態空間模型去估計起迄旅次矩陣並考慮了旅次時間。而這些方法主要是建立了在Kalman過濾器與Gibbs取樣器上。這模型也將擴展到非高斯的觀察誤差。

As origin-destination trip matrices becoming more and more important for many dynamic traffic network control and management analysis, approaches to estimate such matrices from traffic counts have attracted much research interest over the past decade. The dynamic origin-destination estimation approaches are relatively new. Their current status of development and applications are far from being as well recognized as those of static models. In this thesis we provide methods for estimating origin-destination demand pattern in the time domain. For doing this we consider the state space model with travel times to estimate parameters. These techniques rely on Gibbs sampler and Kalman filter. The model will also be extended to include non-Gaussian observation errors.