ISTRC 2026 WEBINAR SERIES - Smart Transportation in Transition
Smart Transportation in Transition is a webinar series by the Israel Smart Transportation Research Center (ISTRC) that presents emerging research, methods, and applications in smart mobility. The webinar series aims to create a forum for discussing how transportation systems are changing in response to emerging technologies, behavioral shifts, sustainability goals, policy needs, and evolving mobility patterns. The series seeks to connect academic research with real-world transportation challenges and to encourage dialogue across disciplines and sectors.
WEBINAR NO. 3:
Error Bounds for Stochastic User Equilibrium Traffic Assignment
DATE & TIME
Nov 16, 2026 | 15:30
TARGET AUDIENCE
Smart Transportation Researchers and Practitioners
SPEAKER
Prof. Stephen D. Boyles
Clyde E. Lee Endowed Professor and Distinguished Teaching Professor in the Fariborz Maseeh Department of Civil, Architectural, and Environmental Engineering
The University of Texas at Austin
FACILITATOR
Dr. Rui Yao
Tenure-track Assistant Professor
Faculty of Civil and Environmental Engineering
Technion – Israel Institute of Technology
DESCRIPTION
Traffic assignment models are used to predict travelers' route choices in a network; stochastic traffic assignment models assume this route choice is based on minimizing the cost, including randomly-distributed perception errors. These models are solved numerically, using iterative algorithms, raising the question of the level of precision needed for practical applications. Past work has developed empirical guidance on the precision levels at which the solution stabilizes, but no mathematical bound on distance to the solution has been known. We present error bounds that do exactly that: at an intermediate point in the computation, we provide upper bounds on how much the path flows, link flows, and other metrics differ from the exact solution. We connect these mathematical bounds to past empirical guidance, and show how the bounds can be used as a pruning rule in bilevel optimization (e.g., model calibration or network design), to focus computation time on the most promising solutions.