Michigan Department of Transportation

While much research has recently focused on generating physics-based adversarial samples, a critical yet often overlooked category originates from physical failures within on-board cameras—components essential to the perception systems of autonomous vehicles. Camera failures, whether due to external stresses causing hardware breakdown or internal component faults, can directly jeopardize the safety and reliability of autonomous driving systems. Firstly, we motivate the study using two separate real-world experiments to showcase that indeed glass failures would cause the detection based neural network models to fail. Secondly, we develop a simulation-based study using the physical process of the glass breakage to create perturbed scenarios, representing a realistic class of physics-based adversarial samples. Using a finite element model (FEM)-based approach, we generate surface cracks on the camera image by applying a stress field defined by particles within a triangular mesh. Lastly, we use physically-based rendering (PBR) techniques to provide realistic visualizations of these physically plausible fractures. To assess the safety implications, we apply the simulated broken glass effects as image filters to two autonomous driving datasets- KITTI and BDD100K- as well as the large-scale image detection dataset MS-COCO. We then evaluate detection failure rates for critical object classes using CNN-based object detection models (YOLOv8 and Faster R-CNN) and a transformer-based architecture with Pyramid Vision Transformers. To further investigate the distributional impact of these visual distortions, we compute the Kullback-Leibler (K-L) divergence between three distinct data distributions, applying various broken glass filters to a custom dataset (captured through a cracked windshield), as well as the KITTI and Kaggle cats and dogs datasets. The K-L divergence analysis suggests that these broken glass filters do not introduce significant distributional shifts. Our goal is to provide a robust, physics-based methodology for generating adversarial samples that reflect real-world camera failures, with the overarching aim of improving the resilience and safety of autonomous driving systems against such physical threats.

Low interfacial thermal conductance often emerges as a primary barrier to effective heat management in advanced nanodevices. This study examines how topological defects affect the interfacial thermal conductance of graphene/SiC lateral heterostructure, utilizing nonequilibrium molecular dynamics simulations. The significant lattice mismatch between graphene and SiC results in a pristine interface that experiences severe strain and structural distortion, ultimately reducing the level of phonon transmission. By introduction of 5|8|5 topological defects, the interfacial deformation is effectively alleviated, thereby improving phonon coupling across the boundary. The results reveal an unconventional increase in interfacial thermal conductance, with the maximal value achieved when three defects are incorporated, representing a 61% improvement compared with the pristine interface. However, an excessive number of defects can lead to a reduction in the thermal conductivity. These findings demonstrate that controlled defect engineering offers a tunable pathway to optimize interfacial heat transport in 2D heterostructures, providing valuable insights for thermal management in nanoscale devices.

Decision-support tools are fundamental to planning road and highway infrastructure maintenance, guiding treatment interventions and their timing. Despite the development of numerous complex decision-support tools, public sector agencies frequently hesitate to adopt them. This paper presents a metamodel framework designed to bridge this gap by matching transportation asset management decision-support tools to specific decision-making contexts or scenarios, thereby promoting use of the most appropriate tools. The framework uses machine learning methods to characterise scenarios and select the most suitable model. The framework is applied to three distinct decision-support tools to determine maintenance policies for pavements and bridges: a comprehensive tool that minimises total costs (agency, user, and disruption) over a planning horizon using Monte Carlo simulation (in which random sampling is used to represent different condition states) and network equilibrium, a simplified fixed policy tool, and a tool of intermediate complexity. Trained and validated using pseudo-data from 400 scenarios across 10 simulated networks, and further tested on a realistic network, the metamodel classifier accurately predicts the most preferred decision-support tool for previously unseen scenarios.

Transportation accessibility remains a critical challenge for visually impaired individuals, constraining their autonomy and societal participation. Although autonomous vehicles (AVs) hold transformative potential for enhancing mobility, prevailing human-machine interfaces (HMIs) frequently neglect the unique interaction requirements of this population. This study investigates the efficacy of a multimodal HMI explicitly designed to facilitate autonomous ridesharing interactions for visually impaired users. Employing a between-subjects experimental design, we evaluated user trust and satisfaction across six core ridesharing functions under three distinct conditions: (1) visually impaired participants with multimodal (audio-visual) feedback, (2) non-visually impaired participants, and (3) visually impaired participants without audio feedback ( N = 24). Our findings demonstrate that audio-enhanced multimodal interfaces bridge the accessibility gap, enabling visually impaired users to attain trust and satisfaction levels statistically comparable to those of non-visually impaired users. Furthermore, the absence of audio feedback significantly degraded navigational confidence, vehicle identification accuracy, and overall user experience ( p < .05). These results theoretically validate the significance of auditory cues in AV HMIs, while empirically confirming design principles for universal accessibility. By providing actionable guidelines for inclusive interface design, this work advances equitable mobility solutions and underscores the imperative of user-centered autonomy in next-generation transportation systems.

Traditional construction estimation relies on historical unit bid prices, which may not reflect actual construction costs due to site-specific conditions, market dynamics, and contractor strategies. However, few studies have systematically examined the full range of factors driving unit price variability, particularly under changing economic conditions. This research thus aims to identify key pricing factors in highway construction, including economic and regional variables, along with the impacts of the COVID-19 pandemic and high inflation. A two-step methodology was used. First, a literature review identified 80 pricing factors spanning macroeconomics, market, regional, client, and project categories. This was followed by a quantitative analysis employing Multivariate Regression, Random Forest, and Ensemble Learning models, applied to 14 years of bid data from the Michigan Department of Transportation. Key findings show that ensemble learning models outperformed other methods in predicting contract-level unit bid prices, achieving a higher explanatory power with an R2 of 0.63. Item quantity and regional spending patterns strongly influence bid prices. For example, larger quantities lower unit prices due to economies of scale. The study highlights the impact of the COVID-19 pandemic on construction costs, driven by supply chain disruptions, labor shortages, and material price inflation.

This study aims to support the inclusive design of autonomous shared rides (ASR) by identifying gaps related to efficient trips and human-machine interaction, specifically for people with Parkinson’s disease (PwPD). In-person interviews were conducted with 20 PwPD, aimed to understand PwPD’s travel experiences, potential user barriers, and needs with regard to an ASR service. During the interview, participants watched short video clips describing five trip segments (proposed by a U.S. Department of Transportation report) of an ASR trip (scenario animations) and responded to questions about these scenarios. Both qualitative (opinions) and quantitative (ranking/rating) data were collected. Results of the Friedman test indicated significant differences in PwPD’s rankings of various travel barriers. Safety and lack of customer service were among the top concerns for PwPD. Qualitative analysis of the interview data further suggested that PwPDs were mainly concerned with the following aspects of ASR: safety (ASR reliability and operation), availability and quality of real-person online customer service and human assistance, user-friendly technology with clear instructions , and accessibility for PwPD with varying levels of mobility, the capability of ASR to deal with emergency situations , and the assistance provided for finding seats and using seat belts . Overall, most PwPD participants ranked safety concern, lack of travel support/customer service , and technology issues as the top three travel barriers for ASR. Among the five trip segments (booking, identification, onboarding, traveling, and exiting), booking was perceived as the most anxiety-provoking segment. These unique data and findings have identified user barriers and needs for ASR, which can guide the design and implementation of future technical solutions to address a broader range of use groups.