ORCID

https://orcid.org/0000-0002-9531-8868

Keywords

Intelligent Transportation System, Intersection, Data-Driven Decision, Automated Traffic Signal Performance Measures (ATSPM), Connected Vehicles

Abstract

Intersections are complex traffic environments where frequent vehicle and pedestrian interactions can lead to safety- and mobility-critical (SMC) conditions. Identifying these conditions promptly is crucial, as delays may result in missed interventions, thereby heightening crash risk, and diminishing operational efficiency. This dissertation introduces a data-driven, real-time decision support framework for the identification and mitigation of SMC conditions using High-Resolution Traffic Controller Data (HTCD) collected via Automated Traffic Signal Performance Measures (ATSPM) system and Connected Vehicle (CV) data from Signal Analytics. First, an end-to-end analytical pipeline was developed to transform HTCD into actionable safety and mobility measures. A Bayesian beta-binomial framework, complemented by non-parametric distributional analysis, was developed to identify pedestrian-related SMC conditions using performance measures from HCTD, enabling proactive, data-driven recommendations for signal control strategies such as Pedestrian Recall, Leading Pedestrian Interval, and No Right-Turn on Red. An interactive dashboard was also developed to visualize performance trends, monitor SMC conditions, and deliver phase-specific interventions, facilitating practical deployment and operational decision-making. Second, Causal Forest–based framework was applied to evaluate the effect of yellow and red-clearance interval adjustments on Red-Light Running (RLR) using HTCD. The model estimated heterogeneous treatment effects and revealed that extending yellow duration by 1 second could reduce red-light running by up to 30 percentage points. Third, a hybrid Transformer-GAN model was developed to predict cycle-level crash likelihood by modeling rare, crash-prone patterns as temporal anomalies using imbalanced HTCD. This model achieved 76% sensitivity, eliminating the need for synthetic oversampling. Fourth, a zone-specific, time-embedded Transformer model was proposed using CV data, as an alternative to HTCD, to predict crash likelihood without fixed infrastructure. Even under low CV penetration (~8%), the model demonstrated strong predictive performance and interpretability, identifying features like control delay and approach speed as key SMC indicators. Collectively, the proposed methods support scalable, sensor-agnostic, and explainable traffic safety interventions, enabling real-time decision support and advancing the goals of Vision Zero through proactive intersection management.

Completion Date

2025

Semester

Summer

Committee Chair

Abdel-Aty, Mohamed

Degree

Doctor of Philosophy (Ph.D.)

College

College of Engineering and Computer Science

Department

Department of Civil, Environmental, and Construction Engineering

Format

PDF

Release Date

8-15-2026

Document Type

Dissertation

Campus Location

Orlando (Main) Campus

Subjects

Intelligent transportation systems--Decision making; Traffic safety--Data processing; Intelligent transportation systems--Safety measures; Traffic engineering--Research; Roads--Interchanges and intersections--Safety measures--Evaluation

Download

Available for download on Saturday, August 15, 2026

Share

COinS
 

Accessibility Statement

This item was created or digitized prior to April 24, 2027, or is a reproduction of legacy media created before that date. It is preserved in its original, unmodified state specifically for research, reference, or historical recordkeeping. In accordance with the ADA Title II Final Rule, the University Libraries provides accessible versions of archival materials upon request. To request an accommodation for this item, please submit an accessibility request form.