An exploration of calibrating activity-based

Parameter tuning of transportation allocation models is essential for travel demand analysis and management. As an extension of traditional traffic allocation, Bounded-Rational Activity-Travel Allocation (BR-ATA) endogenously combines activity-based modeling and traffic allocation to provide high-dimensional selection between high-dimensional selection facets along activity-travel patterns. interdependencies can be captured. The inclusion of multiple episodes of activity participation and restricted rationality behavior expands the selection space and poses challenges for tuning the BR-ATA model. To date, no formulations and solution approaches have been developed for parameter calibration of BR-ATA. To solve this problem, Dong Wang and Feixiong Liao formulated his BR-ATA calibration as an optimization problem and solved it using the simultaneous perturbation stochastic approximation method.

They published their study on January 20, 2023. Communication in transport research (https://doi.org/10.1016/j.commtr.2023.100092).

“We formulate the BR-ATA calibration as an optimization problem, thanks to the multistate supernetwork, and analyze the influence of two additional components on the calibration problem. Considering the time dimension, the BR-ATA calibration problem We also propose a dynamic formulation of Employing a simultaneous perturbation stochastic approximation algorithm to solve the proposed calibration problem Numerical examples are given to graphically adjust activity-based travel demand ,” says Dr. Feixiong Liao, traffic scientist in the Urban Planning and Transportation Group at Eindhoven University of Technology (Netherlands).

BR parameters and activity participation influence calibration

The execution time will be within the range of [0.30, 0.40] Time at which the BR parameter takes different values. Note that ATA calibration issues can take over 2 hours to reach a dead state. Regarding the effect of the number of activities, the execution time decreases as the number of activities increases.

“The execution time in which the BR-related parameters fall within [0.05, 0.2] Relatively stable and shorter than for smaller parameters (i.e. 0.01 or ATA calibration issues). Also, low activity always links flows concentrated in a specific period of time. Link congestion brings more ATP to balance the OD demand. ‘ explains Dong Wang, associate professor at Qingdao University (China).

T.Temporal and Spatial Dimensional Extensions to the BR-ATA Calibration Problem

SPSA takes 8.2 hours to reach the stopping condition for the BR-ATA calibration problem of the Sioux Falls network, and the calibrated demand approximates the prior value. Taking the location of a house as an example, the study shows that the relative error in adjusted demand is 0.01. For the BR-DATA calibration problem of the Sioux Falls network, the run time is 0.92 hours and the number of iterations is 647. All calibrated demands approach the prior value. To further illustrate the scalability on large networks, calibration of the BR-DATA model was performed on the Eastern Massachusetts network. SPSA takes over 10 hours to complete 1000 iterations and the corresponding RMSN (measure of error) is as low as 0.06.

“This result shows that the SPSA algorithm is viable for BR-ATA and BR-DATA calibration problems in large-scale networks,” said Feixiong Liao. “However, large-scale real-world applications require more efficient algorithms,” he adds.

The above research is published in Communications in Transportation Research (COMMTR), a fully open-access journal jointly published by Tsinghua University Press and Elsevier. COMMTR publishes high-quality, peer-reviewed research that represents significant advances for new transportation systems. COMMTR is also one of the first transport journals to mandate a replication package to facilitate the understanding and development of existing knowledge by researchers, practitioners, and the general public. Tsinghua University Press will pay open access fees for all articles published between 2021 and 2025 at its sole discretion.

##

almost Communication in transport research

Communication in transport research publishes high-quality, peer-reviewed research that represents significant advances for emerging transportation systems. Its mission is to provide authors with fair, rapid, and expert peer review, providing readers with insightful theories, impactful advances, and interesting findings. Important general topics, multidisciplinary nature (transport, civil, control, artificial intelligence, social sciences, psychological sciences, medical services, etc.), systems of complex and interrelated systems, strong evidence submissions are welcomed. increase. The strength of the data, forward-looking analysis and predictions for the future, and potentially implementable and available policies/practices. Communication in transport research will be indexed on Scopus 10 months after launch.

Communication in transport research It is a fully open access journal. It is jointly published by Tsinghua University Press and Elsevier, and is jointly sponsored by the National Key Research Institute for Automobile Safety and Energy (Tsinghua University) and China Intelligent Transportation System Association (ITS China). Tsinghua University Press will pay open access fees for all articles published between 2021 and 2025 at its sole discretion.

almost Tsinghua University Press

Founded in 1980 and affiliated with Tsinghua University, Tsinghua University Press (TUP) is China’s leading comprehensive higher education and professional publishing house. Committed to building a top-level global cultural brand, after 41 years of development, TUP has established an excellent management system and corporate structure, covering books, audio, video, electronic products, journals and digital publications. We offer multimedia and multidimensional publications. In addition, TUP was selected as a First-class National Publisher for proactively implementing a strategic transformation from educational publishing to content development and services for teaching and learning, and achieving remarkable results.