|
Activity Number:
|
115
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 9, 2004 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Physical and Engineering Sciences
|
| Abstract - #301669 |
|
Title:
|
Origin-Destination Matrix Estimation from Partially Observed Packet/Vehicle Trajectories
|
|
Author(s):
|
Jaimyoung Kwon*+ and Pravin Varaiya
|
|
Companies:
|
University of California, Berkeley and University of California, Berkeley
|
|
Address:
|
367 Evans Hall #3860, Berkeley, CA, 94720,
|
|
Keywords:
|
Origin-Destination matrix ; method of moment estimator ; network ; partially observed data
|
|
Abstract:
|
Accurate information on Origin-Destination (OD) traffic matrices is critical in engineering tasks for both computer and transportation networks. Since observing OD traffic matrices directly is difficult for large networks, they have been estimated either from survey sampling data or from link counts data using network equilibrium, gravity, or distribution-assignment models. Another data type is partially observed packet or vehicle trajectories. We develop Method of Moment (MM) algorithm for estimating time-varying OD matrices from such data. Estimator errors are calculated either from asymptotic approximation or bootstrap. Simulation study shows that the proposed estimator and standard errors estimates perform well under ideal settings. Recently, such partially observed vehicle trajectories data became available from vehicles equipped with ETC transponders and "readers" installed at various locations in San Francisco Bay Area. The algorithm is also applied to these data to derive hourly OD matrices.
|
