Conference presentation
A Bayesian Monte Carlo approach to model calibration for queuing systems
Jan 2005
Abstract
Calibrating models of queuing processes typically requires the collection of data on the arrival and departure of individual vehicles. In this paper an alternative procedure is described which uses Bayesian Monte Carlo methods to update prior estimates of model parameters using observations of changes in queue length over time. Substantial reductions in parameter uncertainty can be achieved by this calibration procedure. The procedure is most effective at reducing uncertainty in arrival rates when service rates are already well characterized. A number of transportation systems, including toll plazas and inspections stations at ports-of-entry, have well-characterized service capacities, indicating that this method may be appropriate for use with these systems. In general, posterior uncertainties for arrival rates and service rates are higher than for conventional calibration procedures. This procedure is likely to be useful in cases, such as international ports-of-entry, where security and access concerns render the collection of data on individual vehicles infeasible, but abundant information is available on queue lengths.
Metrics
22 File views/ downloads
23 Record Views
Details
- Title
- A Bayesian Monte Carlo approach to model calibration for queuing systems
- Creators
- Patrick L. Gurian - Drexel UniversityFelipe Castro - The University of Texas at El PasoYi-Chang Chiu - The University of Texas at El Paso
- Conference
- Annual Meeting of the Transportation Research Board, 84 (Washington, DC, Jan 2005)
- Resource Type
- Conference presentation
- Language
- English
- Academic Unit
- Civil, Architectural, and Environmental Engineering; College of Engineering
- Other Identifier
- 991014632526504721