Vehicle Detector Deployment Strategies for the Estimation of Network Origin-Destination Matrix

• Main Authors: Chih-Jen Shih, 石智仁
• Other Authors: Shou-Ren Hu
• Format: Others
• Language: zh-TW
• Published: 2007
• Online Access: http://ndltd.ncl.edu.tw/handle/54185389640429534720

碩士 === 國防管理學院 === 國防決策科學研究所 === 95 === Users’ trip Origin-Destination matrix (OD matrix) in a transportation network is one of the crucial components for transportation planning. Traditional means of collecting network OD matrix becomes infeasible due to time and budgetary constraints. As the rapid development of Intelligent Transportation Systems (ITS) and advanced traffic data collection technology, inferring network OD matrix by using link traffic count information obtained through vehicle detector (VD) becomes economically feasible. Nevertheless, it is not practical to expect a full-scale VD deployment plan due to monetary constraint. Therefore, the issue of determining the minimum number of required VDs and their corresponding installation locations that achieve a certain degree of accuracy on OD matrix estimation becomes extraordinarily important. Thus, the purpose of this study is to estimate network OD matrix under partial link traffic count information. To deal with the problem, the present research proposes a new concept, that is the linearly independent vehicle detector deployment strategy by installing VDs on the linearly independent links so that linearly dependent link count information can be easily inferred without losing any crucial information. On the other hand, the present research obtains network OD matrix estimates by the constrained generalized least square (CGLS) method. The OD matrix estimates are obtained by using the full link traffic count information inferred by partial information. However, the GLS method causes errors due to the fewer constrains. Alternatively, the present research imposes another constrain, the user equilibrium (UE) assumption, and formulates the problem by a bi-level programming model. To demonstrate the feasibility of the proposed models, the present research employs three sample networks to conduct the case study and the results indicate that the bi-level approach outperforms the GLS method on the estimation of network OD matrix under partial link traffic count information condition.