Research

This page includes links to research papers and technical reports I've written. Feel free to email (sboyles at austin dot utexas dot edu) if you want to discuss these papers or collaborate on a research idea. You can also contact me if you want a preprint of any of these papers, or if you are having trouble accessing them.

[Welcome] [Preprints] [Books] [Journal Publications] [Book Chapters] [Conference Proceedings] [Technical Reports] [Theses]

For theses written by graduate students, please see the group page.

Welcome

I mainly do research in transportation networks, and in applying optimization techniques to transportation systems. If you don't know me, I've chosen the following papers to represent the type of work I do. I'm particularly proud of each of these for some reason or another - they are not all among my most cited work, but they express my research personality well.

  1. Boyles, S. D., N. E. Lownes, and A. Unnikrishnan. (2021) Transportation Network Analysis, Volume I: Static and Dynamic Traffic Assignment.

    This is an open-access textbook I have been writing for several years, based on the graduate courses I teach. It's currently in an open beta phase, so any comments or suggestions you can provide are very welcome!

  2. Boyles, S. D., and N. Ruiz Juri. (2019) Queue spillback and demand uncertainty in dynamic network loading. Transportation Research Record 2673, 38-48. Finalist, 2019 Stella Dafermos Award.

    Transportation models tend to get more and more complex over time. This paper is a bit of a "minority report" in arguing that simplicity is a virtue: in traffic flow, queue spillback makes a model more "realistic" but also more subject to error. If you don't know the model inputs well, you're better off using a simpler model without spillback.

  3. Jafari, E., V. Pandey, and S. D. Boyles. (2017) A decomposition approach to the static traffic assignment problem. Transportation Research Part B 105, 270-296.

    This paper shows how the static traffic assignment problem can be solved by partitioning a network into geographic regions. Smaller assignments can be performed on these regions, alternating with "master" iterations on a simplified model of the full network. We prove that this method always converges to the correct solution in the full network. If the network has a clear partitioning (like a river bisecting the network), run times are substantially reduced. We are currently looking at how best to partition networks when there is no obvious way to do so.

  4. Boyles, S. D., S. Tang, and A. Unnikrishnan. (2015) Parking search equilibrium on a network. Transportation Research Part B 81, 390-409.

    A surprising amount of congestion is due to drivers searching for parking. This paper presents a model for such congestion: drivers search for parking based on their experience of where available spaces are likely to be found. But the likelihood of finding parking at any spot depends on everybody else's search strategies. The resulting model has some neat mathematical features; it is a network flow problem with nonlinear conservation equations. When we wrote this, I had not seen such a structure before. Since then, I've learned there are parallels with telecommunication systems and queueing networks.

  5. Khani, A. and S. D. Boyles. (2015) An exact algorithm for the mean-standard deviation shortest path problem. Transportation Research Part B 81, 252-266.

    The classical shortest path problem is easy. Minimizing mean travel time in a stochastic network is also easy, and so is minimizing a weighted sum of mean travel time and variance. Minimizing a weighted sum of mean travel time and standard deviation is harder (indeed NP-hard), because the square root breaks Bellman's subpath optimality principle. However, the mean-variance and mean-standard deviation problems are closely related, and we show that the efficient frontier of the latter is a subset of the efficient frontier of the former. This leads to a nice parametric search algorithm based on reoptimizing previously-found solutions. (There is a more recent paper by Zhang et al. showing even better results with an enhanced parametric search.)

  6. Boyles, S. D. (2012) Bush-based sensitivity analysis for approximating subnetwork diversion. Transportation Research Part B 46, 139-155.

    This paper was the foundation of my NSF CAREER grant, suggesting that "subnetwork" models should not be treated as independent of the full network. Rather, augmenting the subnetwork with a simplified representation of the full network improves the consistency of the resulting solution. I'm particularly fond of the solution methods proposed here: one draws connections to Kirchhoff's laws in electric circuits and reductions from graph theory, and the other makes use of some fun linear algebra tricks. We later found better ways to solve this problem in practice (see this follow-up paper with Ehsan Jafari) but I still like the connections this paper made to other domains.

Preprints

The following papers are currently under review or preparation. I am working on uploading preprints; for now if no link is provided, contact me if you want a copy.

  • Gokalp, C., S. D. Boyles, and A. Unnikrishnan. Mean-standard deviation model for minimum cost flow problem. In review, Networks.
  • Zhu, T., S. D. Boyles, and A. Unnikrishnan. Two-stage robust facility location problem with drones. In review, Transportation Research Part C.

Books

Journal Publications

Book Chapters

Conference Proceedings

Technical Reports

Theses

Back to the main page.