Haitao Liao

Unmanned aerial vehicle (UAV) roof inspections are often limited by restricted flight times and the inefficiency of uniform coverage strategies. This study introduces an adaptive weighted coverage planning framework that prioritizes structurally informative regions using visual cues, bypassing the need for explicit defect annotations. The framework generates a grid-level informativeness map by fusing multiscale keypoint density, seam detection, color consistency, and illumination reliability. To optimize deployment, high-priority regions are aggregated into inspection blocks via a greedy region-merging algorithm, effectively decoupling region selection from path planning. A nearest neighbor (NN) planner + 2-opt refinement is then utilized for real-time trajectory generation and compared against randomized approximate neighbor + 2-opt and simulated annealing methods. Experimental results on high-resolution UAV imagery of concrete/cement roofs demonstrate that the framework concentrates inspection efforts on critical structural areas while suppressing redundant, homogeneous regions. Quantitative analysis shows that NN + 2-opt provides an optimal balance between trajectory quality and computational efficiency. This modular approach offers a scalable, interpretable solution for autonomous infrastructure monitoring, directly addressing the operational constraints of limited UAV flight time and the diagnostic inefficiency of uniform coverage strategies in large-scale facility inspection.

U.S. freight transportation faces increasing highway congestion and greenhouse gas (GHG) emissions. Container-on-barge (COB) integration offers an alternative but remains underutilized in the Lower Mississippi River (LMR), with benefits dependent on operating conditions. This study develops a discrete-event simulation model to evaluate the impacts of COB integration on system time, fuel costs, and GHG emissions for the LMR. Four scenarios examine a truck-only baseline, progressive COB adoption, seasonal water-level variability, and cargo time sensitivity. Under progressive adoption, a 10% COB share reduces fuel cost, CO$_2$ emissions, and system time by 9.15%, 8.78%, and 19.83%, respectively, compared to the baseline. Low-water conditions limit barge capacity, while assigning time-sensitive freight to trucks supports deadline compliance, highlighting trade-offs between emissions reduction and service requirements as COB adoption increases. These findings underscore the importance of balancing emissions reduction and fuel savings with operational constraints to support sustainable COB integration in the LMR corridor.

Floods create serious obstacles for land transportation, especially critical roads needed for emergency actions when every minute counts. In such situations, predicting vehicle speeds on critical roads is essential for road usability assessment and planning for efficient emergency response. Although quite a few models for vehicle speed vs. water depth have been developed using controlled experiments, these models often fail to capture how drivers behave over the period of flood. In this study, we propose a video-based approach to develop probabilistic models that account for time-dependent vehicle speeds by analyzing real-world video footage of vehicles navigating floodwaters. The approach combines state-of-the-art YOLOv11 object detection, Kalman filter-based tracking, and perspective-aware speed computation to extract vehicle speeds, even in complex conditions involving occlusion, camera angle variation, jitter, and adverse weather conditions. Based on time-dependent vehicle speed data, Weibull distribution-based models reflecting how drivers behave on flooded roads with diverse traffic scenes are developed to predict vehicle speeds over time. The proposed framework for data collection and probabilistic model development is validated using real-world surveillance and YouTube data. The results show that the video-based approach can achieve high accuracy in collecting vehicle speeds, with a mean absolute error below 12 mph, and can operate in real time, making it suitable for scalable urban deployment. Moreover, the proposed probabilistic models provide realistic and reliable tools for predicting vehicle speeds over time, which assist emergency services in predicting road and traffic conditions to ensure driving safety and efficient emergency response during and after a flood.

1. Optimisation is one of the most important tasks in science and engineering applications. To handle real-world optimisation problems, numerous deterministic and stochastic methods have been devel...

Critical infrastructures, such as transportation, power, water, and communication networks, are vital to a nation's economic and productive activities. Since these infrastructures are often vulnerable to failures caused by aging, natural hazards, and human-induced disturbances, it is crucial to efficiently assess their reliability over time and make necessary improvements against failures. As one of the most important measures for network reliability, all-terminal network reliability is defined as the probability that all nodes in a network are fully connected. However, as the numbers of edges and nodes increase, calculating all-terminal network reliability using an exact approach is extremely time consuming and sometimes becomes infeasible within a limited time. The technical barrier is more significant when the goal is to improve and/or maintain the reliability of an existing infrastructure network, as the reliability of the network with a time-dependent topology must be constantly evaluated and optimized sequentially. To overcome these challenges, we propose the use of Graph Neural Networks (GNNs) to efficiently estimate all-terminal network reliability and provide a sequential network optimization framework by taking advantage of Deep Reinforcement Learning (DRL). Specifically, given an existing network and available resources, the goal is to determine the best sequence of maintenance, repair, and topology changes to maximize the network's all-terminal reliability over a finite time horizon. Our numerical experiments show that the GNN model outperforms those alternatives (e.g., Deep Neural Network, Convolutional Neural Network, Monte Carlo (MC) simulation) reported in previous studies in estimating all-terminal network reliability, and the DRL framework can efficiently suggest a sequence of maintenance or improvement actions to assist decision makers in achieving their reliability goals for infrastructure systems.