On the Use of Abundant Road Speed Data for Travel Demand Calibration of Urban Traffic Simulators

TL;DR

This work tackles origin-destination travel-demand calibration for urban traffic microsimulation using abundant segment speed data. It introduces a physics-informed metamodel that blends a fundamental diagram–based analytical relation with a linear residual model to approximate the simulation loss, enabling gradient-based optimization. The problem is formulated as a bound-constrained, simulation-based optimization of $f(x)=\frac{1}{|\mathcal{I}|}\sum_{i\in\mathcal{I}} w_i (v_i^{GT}-E[v_i(x,u_1;u_2)])^2$ and solved iteratively by updating the metamodel with new simulations. Applied to a Salt Lake City network, the method achieves substantial calibration improvements over SPSA, with larger gains when more GT-speed data are available, and demonstrates superior compute efficiency by requiring fewer simulation evaluations.

Abstract

This work develops a compute-efficient algorithm to tackle a fundamental problem in transportation: that of urban travel demand estimation. It focuses on the calibration of origin-destination travel demand input parameters for high-resolution traffic simulation models. It considers the use of abundant traffic road speed data. The travel demand calibration problem is formulated as a continuous, high-dimensional, simulation-based optimization (SO) problem with bound constraints. There is a lack of compute efficient algorithms to tackle this problem. We propose the use of an SO algorithm that relies on an efficient, analytical, differentiable, physics-based traffic model, known as a metamodel or surrogate model. We formulate a metamodel that enables the use of road speed data. Tests are performed on a Salt Lake City network. We study how the amount of data, as well as the congestion levels, impact both in-sample and out-of-sample performance. The proposed method outperforms the benchmark for both in-sample and out-of-sample performance by 84.4% and 72.2% in terms of speeds and counts, respectively. Most importantly, the proposed method yields the highest compute efficiency, identifying solutions with good performance within few simulation function evaluations (i.e., with small samples).

Figures (1)

• Figure 1: Simulated vs. GT speeds for SPSA (left) and metamodel (right) calibrated demands. Top, middle, and bottom rows show results for calibration with 331, 456, and 2,282 segments, respectively. The $x=y$ line is shown for reference.