Title: 以路段交通量推估起迄分佈矩陣之研究
The Study of Origin-Destination Matrix Estimation From Link Traffic Count Data
Authors: 施世欽
Shih-Chin Shih
卓訓榮
Hsun-Jung Cho
土木工程學系
Keywords: 起迄分佈矩陣;習慣領域;Origin Destination Matrix;Habitual Domains
Issue Date: 1994
Abstract: 起迄分佈矩陣在運輸系統規劃上是一項重要的資料。傳統獲得起迄分佈矩
陣的方法,如家戶訪問法、車輛牌照登錄法等,調查的成本昂貴且困難並
有相當大的誤差,因此有許多研究者嘗試由交通量資料來推估起迄分佈矩
陣,以代替傳統的調查方法。過去推估起迄分佈矩陣的研究,在分析用路
人的行為大部分皆以使用者均衡為準則,但其條件過於嚴格,另外有些研
究的路段成本函數並不適用於研究地區。因此本研究的目的在於探討研究
地區的成本函數及發展較符合實際用路人行為的推估方法。本研究透過實
際調查,蒐集研究地區路段的旅行時間及流量,利用 SPSS校估出路段成
本函數,並且在研究地區路網對行駛車輛連續調查五天,發現用路人並不
一定掌握路網的充分資訊,未必選擇最短的路徑來行駛,且有使用習慣路
徑的現象。因此本研究以習慣領域理論做為用路人選擇路徑的行為基礎,
即認為用路人由於過去不斷的行駛經驗,而形成本身對路徑選擇的習慣領
域,在選擇路徑時會以此習慣領域做為決策的依據,也就是每一個用路人
皆會行駛其習慣走的路徑。路徑習慣使用的機率採用具有相互比較關係的
羅吉特模式形式,利用最小平方法構建推估起迄分佈矩陣的模式,以修正
最陡坡降法及黃金切割法求解,並以實際調查的資料進行測試,以瞭解模
式的可行性。
Origin-Destination(O-D) matrix is an important data in ur- ban
transportation planning. Conventional method such as home
interview surveys for obtaining O-D matrix in urban areas are
prone to considerable sampling error and become prohibitively
costly. For these reasons several researchers try to develop
methods for estimating O-D matrix from link traffic count data
. Most previous research works on O-D matrix estimation conce-
ntrated on rule of user-equilibrium whiich analyzes driver's
behavior, but its condition is too stringent.Therefore, the p-
urpose of this thesis to develop an estimation method which is
more correspond real driver's behavior. For vehicles continue
to survey at study area for five da- ys.We discover which
drivers are uncertain clutch network suf- ficient information,
uncertainly selects shortest path and it has the phenomenon of
use habitual path. This thesis with hab- itual domains makes
drivers select path basis. Assume every d- river be able select
habitual path. Path habitual drive proba- bility adopts the
form of logit model. Use the method of leas- t square to
construct the model of O-D matrix estimation.
URI: http://140.113.39.130/cdrfb3/record/nctu/#NT830015050
http://hdl.handle.net/11536/58743
Appears in Collections:Thesis