United States Department of Transportation
i
Terminal Location Planning in Intermodal Transportation With Bayesian Inference Method
-
2014-01-01
-
[PDF-564.75 KB]
Details:
-
Creators:
-
Corporate Creators:
-
Corporate Contributors:
-
Subject/TRT Terms:
-
Resource Type:
-
Geographical Coverage:
-
Corporate Publisher:
-
Abstract:In this project, we consider the planning of terminal locations for intermodal transportation systems. For a given number of potential terminals and coexisted multiple service pairs, we find the set of appropriate terminals and their locations that provide the economically most efficient intermodal operation. The first part of this project is to develop a two-layer Markov Chain Monte Carlo (MCMC)-based method to implement the terminal location planning. The lower-layer is an optimal routing algorithm for all service pairs that considers both efficiency and fairness for a given planning. The upper-layer is a planning algorithm based on MCMC with a stationary distribution mapped from the transportation cost function. This method has shown, as tested in various network scenario, better performance than a recently developed method using a greedy randomized adaptive search procedure together with a heuristic search procedure (GRASP-HEP). In the second part of the project, we bring the probabilistic nature in transportation networks into consideration. Estimates of traffic needing to use the network, capacity of terminals and costs of using portions of the network vary time to time. Effects of these variations have not been previously studied in the literature. We characterize the uncertainty of the system parameters with probability density functions (PDFs) based on prior information, while map the cost function into a likelihood function. Then, the design problem can be converted into a Bayesian inference problem of finding parameter set solutions with high posteriori probability that is proportional to the product of the prior PDF and the likelihood. We have developed theoretic methods for uniform sampling multi-dimensional simplex volume and implemented the Nested Sampling method to rank solutions based on their evidence values.
-
Format:
-
Alternate URL:
-
Funding:
-
Collection(s):
-
Main Document Checksum:
-
Download URL:
-
File Type:
