Campus Parking Permit Model

The question

The University of Arkansas has 14,632 parking spaces and sold 21,124 annual car permits in fiscal 2025, about 1.44 for every space. In a 2024 parking study with more than 5,000 survey responses, 87.4% said they were unsatisfied. A 2025 university survey found the top complaints were congestion, the distance from parking to where people are going, and finding a space, and the university listed a zone-based permit system among the changes it's considering.

Our question: if you could redraw which permit types can use each lot, which layout would shorten the walk from parking to buildings the most, and what would it cost each group of drivers?

Today's layout

Commuter and Economy (18 lots)
Faculty/Staff (24 lots)
Resident, reserved, garages, pay and other (68)

All 110 lots and garages on the university's campus map, colored by the permit each serves today. Hover a lot for its name and estimated size. Commuter, Economy and Faculty/Staff lots are the ones our model can redesignate. Resident and reserved lots stay fixed.

Show the numbers

Designation today	Lots
Commuter	17
Economy (Lot 500)	1
Faculty/Staff	24
Resident	28
Reserved	22
Garages	4
Hourly, ADA, short-term, visitor, relocate and other	14
Total	110

The model

The model makes two kinds of decisions at once: which permit type each lot serves (Commuter, Faculty/Staff or Economy; resident and reserved lots stay as they are), and how many drivers from each group park in each lot. It minimizes the total walking distance from lots to buildings, weighted by group.

SetsG groups of drivers, L lots, P permit typesgroups by permit and destination

Variablesy_lp ∈ {0, 1}lot l serves permit type p

x_gl ≥ 0drivers from group g who park in lot l

MinimizeΣ_g Σ_l w_g d_gl x_glweighted walking distance

Subject toΣ_l x_gl = n_gevery group gets a place to park

Σ_g x_gl ≤ r s_lno lot overfills; r allows for permit holders who aren't on campus at once

x_gl ≤ n_g Σ_{p ∈ P(g)} y_lpdrivers only park where their permit is allowed

Σ_p y_lp = 1one permit type per lot

ExtensionΣ_{l ∈ L(a)} Σ_g x_gl ≤ C_arush-hour capacity of each access road a

Permits don't map one-to-one onto lots: some lots accept several permit types, so the model works from a compatibility table built from the university's parking rules. The traffic extension keeps the model from shortening walks by sending every driver down the same road.

Open questions we're working through

The Economy problem. Economy lots accept almost every permit, so a naive model would label every lot Economy. It needs a rule that stops that.
One oversell ratio or several. A single ratio of 1.44 probably doesn't fit students and faculty alike, since they're on campus at different rates. We'll know more when lot-usage data arrives.

The data

No one publishes how many spaces each lot has, so we built the dataset from public sources:

110 lots and garages (4 of them garages) from the university's interactive campus map, with outlines and permit rules.
Space estimates from each lot's paved area, calibrated so the surface lots add up to the official 10,181 spaces. The garages hold 4,451 more, from the university's 2023 garage study.
Permits sold by type from the Transit & Parking fiscal 2025 annual report.
Traffic: daily counts from 103 ArDOT stations around campus going back to 1986, and live TomTom road speeds at 12 points, logged every 30 minutes since September 28.

My part

Damir and I built this project together rather than splitting it up: we assembled the dataset, set up the traffic logger and worked out the model side by side.

We also went after the data we couldn't download. I wrote our request for lot-usage data to Transit & Parking and asked the Associated Student Government president for introductions. Damir and I then met with ASG's Director of Transit and Parking on September 30 with a project brief and a list of questions for Parking staff. That request reached Transit & Parking, who are now pulling lot-level data for us.

We'll also check our AMPL results against an AI-written prototype of the model. The model we hand in is ours, written in AMPL and explained in our own words.

What's next

The model and results are due December 3. The experiments we've planned:

Today's layout against the optimized one
A trade-off curve between student and faculty/staff walking
How sensitive the answer is to the space estimates and the oversell ratio
A what-if for the garage the university plans to open on Razorback Road in 2029

The results, a before-and-after map and the walking time saved will go here in December.

Team

Damir Ashinov: Damir's ePortfolio

A student project, not an official University of Arkansas study.

Campus parking permit model