Transportation agencies receive large volumes of free-form public comments describing infrastructure conditions, safety concerns, and service issues. These comments are often processed manually for downstream operational actions, which is time-consuming, inconsistent across reviewers, and difficult to scale, thereby limiting their value for operational decision-making. This study presents a machine learning and Large Language Model (LLM) framework for automated triage of free-form public comments, assigning each report to a three-level hierarchical taxonomy consisting of Category, Subcategory, and Final Decision. The proposed framework uses agency historical data together with retrieval-based evidence, where semantically similar past comments are provided to the LLM as contextual support to better align predictions with agency-specific labeling practices. The framework was evaluated using TF-IDF with Logistic Regression, TF-IDF with Linear SVM, embedding-based kNN with cosine similarity, few-shot LLM prompting, and retrieval-based LLM prompting. Results show that retrieval-based prompting achieved the best overall performance, with the highest accuracy at both the Category and Subcategory levels. At the Final Decision level, retrieval-based prompting slightly outperformed kNN, while few-shot prompting performed worse. Error analysis showed that many misclassifications were semantically plausible alternatives, reflecting the overlap across infrastructure-related complaint categories. When a second candidate label was allowed, further improving performance. Latency analysis also indicated that the framework can process more than 2000 comments in under 30 min, supporting faster and more consistent agency workflows.

This study supports the City of Pittsburgh’s SmartPGH initiative by evaluating advanced pedestrian signal strategies designed to enhance safety and multimodal mobility on high-demand urban corridors. The objective is to identify the best trade-offs between pedestrian timing treatments and signal control modes (actuated free versus coordinated) for varying pedestrian and vehicular demand throughout the day. While prior research has implemented clearance extension logic in simple midblock environments, this study advances the field by applying and testing pedestrian strategies in complex, multiphase intersections using real-world controller logic and time-of-day profiles. The evaluated strategies include pedestrian protection logic (PPL), passive pedestrian detection (PPD), dynamic pedestrian-exclusive phases (DEP), and pedestrian recall (PR). A novel contribution is the implementation of PPL within a fully actuated controller, which selectively delays conflicting vehicle phases to protect pedestrians still crossing, without interrupting concurrent vehicle movements. Evaluation is conducted through a software-in-the-loop simulation framework that integrates microsimulation with Maxtime signal controllers. Results demonstrate that PPD scenarios consistently offer the lowest person delay across a.m., midday, and p.m. periods while maintaining strong vehicle and bus performance. The addition of PPL and DEP further improves pedestrian service, particularly under high pedestrian volumes, with moderate trade-offs in vehicle delay. These findings provide actionable insights for transportation agencies seeking to implement intelligent pedestrian signal strategies as part of broader smart city goals to promote safety and efficiency.

Intersections present considerable safety risks for bicyclists, where road space shared with vehicles frequently triggers conflicts. Although connected technologies promise safety benefits, they may trigger additional stress and adverse behavioral changes, making it important to investigate their impact on bicyclist–vehicle conflicts. This study introduces a novel experimental framework that integrates a high-fidelity digital twin of a one-way signalized arterial (Delaware Avenue in Newark, DE, United States) with a dual human-in-the-loop simulator, allowing real-time interaction between a human-driven vehicle and a human-operated bicycle. The digital twin combines CARLA, used for immersive participant simulation, and PTV Vissim, which replicates realistic background traffic conditions. The first step was to identify potential conflict situations by considering traditional surrogate safety measures and validating them as real conflicts with self-reported feedback and physiological stress responses. Results confirmed that the experimental setup produced events meeting both kinematic and stress-based conflict criteria. The second stage focused on how connected technologies influence user experience, specifically whether collision warning information evokes additional stress, given its novelty and limited adoption in real-world contexts. Finally, we examined behavioral changes by analyzing approach and postconflict speeds, revealing how connectivity affects speed choices in conflicting situations, with drivers increasing approach speeds and bicyclists becoming more cautious. This research demonstrates the value of combining digital twins, dual human-in-the-loop simulators, and physiological sensing for studying connected safety technologies. The findings provide new insights into how connectivity influences user stress and behavior, helping the design and evaluation of future connected multimodal traffic systems.