作者: Klayut Jintanakul
DOI:
关键词: Sampling (statistics) 、 Dynamic demand 、 Computer science 、 Microsimulation 、 Mathematical model 、 Mathematical optimization 、 Simulation 、 Monte Carlo method 、 Traffic engineering 、 Transportation planning 、 Traffic simulation
摘要: Author(s): Jintanakul, Klayut | Abstract: A spectrum of traffic engineering and modern transportation planning problems requires the knowledge underlying trip pattern, commonly represented by dynamic Origin- Destination (OD) tables. In view fact that direct survey pattern is technically problematic economically infeasible, there have been a great number methods proposed in literature for updating existing OD tables from counts and/or other data sources. Unfortunately, remain several common theoretical practical aspects which impact estimation accuracy limit use these most real-world applications. This dissertation itemizes examines critical issues. Then, presents developments, evaluations, applications two new frameworks intended to be used with current near-future data, respectively.The first framework offers systematic procedure preparing demand inputs microscopic simulation under an module based solely on counts. Under this framework, traditional model augmented filter step, captures important spatial-temporal characteristics route patterns within large surrounding network, improve flow estimates entering leaving final network. bounded solution algorithm solving problem are also proposed.The second utilizes additional information small probe samples collected over multiple days. There steps framework. The step includes suite empirical hierarchical Bayesian models estimating time dependent travel distributions, destination fractions, fractions data. These provide multi-level posterior parameters tend moderate extreme toward overall mean magnitude depending their precision, thus overcoming due non-uniform (over space) sampling rates. involves construction initial tables, route-link via Monte Carlo simulation, using formulation can take into account stochastic properties assignment matrix.