About
Motivation and Problem Context
Understanding the mechanisms governing urban movement remains a central problem across transportation systems modeling and regional planning. While traditional travel behavior scholarship has long focused on characterizing how individuals navigate multimodal networks, translating empirical mobility records into actionable predictive models and public safety interventions poses distinct technical challenges. These challenges manifest across two critical domains:
- Resolving travel mode choices from rich sociodemographic and spatial covariates, as these choices are closely related to Houstonians’ daily commutes and other activities undertaken to meet their needs, and relatedly
- Isolating high-risk transportation corridors from observational collision data to formulate safety countermeasures.
We propose The Houston Mobility Games, a student competition designed to address these challenges through two complementary tracks. Rather than evaluating models in the abstract, this initiative bridges interpretable machine learning and evidence-based policy analysis by anchoring both tracks in localized, real-world data.
- Track A is a leaderboard-scored prediction contest: students predict how students at Houston-area campuses get to campus (i.e., which travel mode they use), and analyze which factors matter most in that that decision.
- Track B is a traffic safety analysis competition: students identify Houston's most dangerous corridors from crash records and propose countermeasures. Students may enter either or both.
Centennial Impact
As UH marks its centennial, the Houston Mobility Games brings together students across disciplines to address mobility and safety challenges affecting the university and the wider Houston region. Combining travel behavior research, predictive machine learning, and spatial analytics, the competition encourages students to connect technical findings with an understanding of people’s transportation needs.
Participants will gain experience in reproducible analysis, model interpretation, and translating evidence into recommendations for safer and more accessible transportation. Their findings could inform campus mobility planning and regional safety discussions, while selected projects could serve as teaching materials and starting points for future student research, extending the competition’s educational value beyond the centennial year.
Data Infrastructure and Track Structure
Both competition tracks leverage fully acquired, cleaned, and profiled datasets, eliminating operational licensing dependencies:
Track A (Mode Choice Inference)
Built on the 2025 Houston-Galveston Area Council (H-GAC) University Travel Survey (N = 6,523 respondents across 14 regional campuses, including 2,739 UH students). The task is framed over 5,518 validated records collapsed into seven discrete access modes:
- Drive Alone (57.0%)
- Dropped Off (20.4%)
- Transit (7.9%)
- Active Transportation (5.4%)
- Carpool (4.0%)
- Campus Shuttle (3.2%)
- Ride-Hail (2.0%)
With a majority-class benchmark of 57.0%, evaluation tests whether models capture non-majority transit and active modes. Features include socioeconomic indicators, geocoded residential census blocks, vehicle availability, and transit pass status.
Track B (Spatiotemporal Crash Modeling):
Built on multi-year TxDOT CRIS collision records for Harris County. Teams identify corridors bearing a disproportionate share of serious injuries and fatalities (KABCO scale), select a focal segment, and quantify countermeasure safety benefits while accounting for traffic exposure and baseline assumptions.
Deliverables & Scoring
Probabilistic Scoring and Rigor (Track A)
Submissions require class probability distributions evaluated against held-out test splits. Performance tracking utilizes a public/private partition to penalize leaderboard overfitting. In addition to top-1 accuracy and separability metrics, teams submit a two-page technical report documenting feature engineering, cross-validation protocols, model architectures, and an interpretation of key predictors and their contributions to travel mode predictions.
Countermeasure Justification and Exposure Control (Track B)
Teams deliver a six-page analytical memorandum directed to municipal stakeholders, accompanied by reproducible code and high-resolution spatial maps. Judging panels evaluate data integrity, severity weighting schemes, explicit accounting for baseline traffic exposure, and transparency regarding analytical limitations
Summary
| Track A — Mode Choice Inference | Track B — Spatiotemporal Crash Analytics | |
|---|---|---|
| Primary Objective | Multiclass mode choice estimation | Spatial risk modeling & countermeasure evaluation |
| Target Audience | Computer science, math, statistics, engineering, data science | Engineering, planning, architecture, public health, policy, business |
| Team size | 1–3 | 2–4 |
| Scoring | 70% automated metric, 30% written model report | 100% expert rubric & Final Presentation |
| Primary Output | Probabilistic submission & validation pipeline | 6-page municipal policy memo & code repo |
Contest Organizers
| Lead organizer | Ekin Ugurel, Assistant Professor, Industrial and Systems Engineering, University of Houston — eugurel@central.uh.edu |
|---|---|
| Co-organizer | Kailai Wang, Assistant Professor, Industrial and Systems Engineering, University of Houston — kwang43@central.uh.edu |
Judges & Agency Partners
- City of Houston: Planning & Development, which runs Vision Zero, and Houston Public Works for traffic operations.
- TxDOT Houston District traffic safety staff. They own the state roads and the CRIS data, and they know Crash Modification Factors well.
- Harris County engineering or precinct staff.
- Houston METRO (Gino)
- H-GAC's regional safety planning staff. (Gino/Kailai)
- Practitioners: the Houston-area ITE and WTS chapters, and consulting firms that write safety action plans. They make good judges and possible prize or internship sponsors
- Greater Houston Partnership (GHP)
Competition Schedule
| Date | Milestone | Description |
|---|---|---|
| Oct 08, 2026 | Registration Opens | Individuals and incomplete teams can sign up. |
| Oct–Nov 2026 | Judge & Agency Partner Recruitment | Ekin, Kailai, and Gino recruit judges and agency partners. |
| Jan 29, 2027 | Team Formation Mixer & Prep Workshop | Formal rules, rubrics, and data dictionaries released. Light snacks and refreshments provided. |
| Feb 5 | Team Registration Closes | Deadline for team registration. |
| Feb 19 (Friday) | Competition Kickoff | Data released. Agency partners frame the Track B problem. |
| Feb 19–28 | Competition Sprint | Two weekends of competition, with midweek office hours. |
| Mar 1–10 | Submission Judging | Judges review Track A reports and Track B memos. |
| Mar 12 | Finalists Announced | Anticipated selection: top 5–6 teams in Track B and top 3 in Track A. |
| Mar 15–21 | Spring Break | No competition activities. |
| Mar 22–Apr 2 | Track B Finalist Feedback | Each finalist receives one 30-minute feedback session with an agency partner. |
| Apr 7 | Finalist Slides Due | Deadline for finalist presentation slides. |
| Apr 9 | Finals | Track B presentations to the panel and short talks from the Track A top 3. |
| Apr 20 | Awards Ceremony | Winners announced and prizes awarded. |
Ready to Register Your Team?
Ensure your team meets all qualifications.
- All team members are eligible UH main campus students for Fall 2026.
- Designated team leader with UH Cougarnet ID and email address.
- Unique team name confirmed.
- Understanding of E10+ rating, AI, copyright, and content rules.
Team Registration
Contact Us
If you have any questions about the contest, please reach out to Kailai Wang, Associate Professor at kwang43@central.uh.edu.