Autonomous construction framework for crane control with enhanced soft actor–critic algorithm and real‐time progress monitoring

Computer-Aided Civil and Infrastructure Engineering, Год журнала: 2025, Номер unknown

Опубликована: Янв. 30, 2025

Abstract With the shortage of skilled labors, there is an increasing demand for automation in construction industry. This study presents autonomous framework crane control with enhanced soft actor–critic (SAC‐E) algorithm and real‐time progress monitoring. SAC‐E a novel reinforcement learning superior speed training stability lifting path planning. In addition, robotic kinematics are implemented to ensure that can autonomously execute path. Last, hardware communication interfaces between robot operating system building information modeling (BIM) developed The performance proposed was demonstrated using robotized mobile stack concrete retaining blocks. results show be effectively used block update BIM platform.

Язык: Английский

---

Efficient 3D robotic mapping and navigation method in complex construction environments

Computer-Aided Civil and Infrastructure Engineering, Год журнала: 2024, Номер unknown

Опубликована: Окт. 9, 2024

Abstract Recent advancements in construction robotics have significantly transformed the industry by delivering safer and more efficient solutions for handling complex hazardous tasks. Despite these innovations, ensuring safe robotic navigation intricate indoor environments, such as attics, remains a significant challenge. This study introduces robust 3‐dimensional (3D) mapping method specifically tailored environments. Utilizing light detection ranging, simultaneous localization mapping, neural networks, this generates precise 3D maps. It also combines grid‐based pathfinding with deep reinforcement learning to enhance obstacle avoidance dynamic settings. An evaluation conducted simulated attic environment—characterized various truss structures continuously changing obstacles—affirms method's efficacy. Compared established benchmarks, not only achieves over 95% accuracy but improves 10% boosts both efficiency safety margins 30%.

Язык: Английский

---

Traffic estimation in work zones using a custom regression model and data augmentation

Computer-Aided Civil and Infrastructure Engineering, Год журнала: 2025, Номер unknown

Опубликована: Янв. 5, 2025

Abstract Accurately estimating traffic volumes in construction work zones is crucial for effective management. However, one of the key challenges transportation agencies face limited coverage continuous count station (CCS) sensors, which are often sparsely located and may not be positioned directly on roads where present. This spatial limitation leads to gaps data, making accurate volume estimation difficult. Addressing this, our study utilized a custom regularized model variational autoencoders (VAE) generate synthetic data that improves estimations these challenging areas. The proposed method only bridges between sparse CCS sensors but also outperforms several benchmark models, as measured by mean absolute percentage error, root square error. Moreover, effectiveness VAE‐augmented models enhancing precision accuracy further underscores benefits integrating into traffic‐modeling approaches. These findings highlight potential approach enhance assist informed decisions

Язык: Английский

---

Hybrid deep learning model for predicting failure properties of asphalt binder from fracture surface images

Computer-Aided Civil and Infrastructure Engineering, Год журнала: 2025, Номер unknown

Опубликована: Фев. 20, 2025

Abstract Cracking impacts asphalt concrete durability primarily due to cohesive binder failures. The poker chip test has recently been introduced better characterize the cracking potential of binders by fracturing a specimen in realistic stress state thin film. However, broader adoption faces challenges high instrumentation costs for measuring load and displacement. This paper presents validates deep learning model that predicts ductility tensile strength from posttest images fractured surfaces, with extensions simplified instrumentation. hybrid model, named PCNet, integrates custom lightweight convolutional neural network (CNN) developed capture local features (e.g., edges, boundaries, contours) within fracture cavities Swin Transformer models global contextual dependencies. A bidirectional cross‐attention fusion module is designed facilitate mutual information exchange between CNN transformer branches. fused are then processed fully connected (FCN) predict indices derived test. proposed demonstrates predictive accuracy across range configurations, achieving an 0.966 mean absolute percentage error (MAPE) 12.95% predicting ductility, while also attaining 0.947 MAPE 9.15% strength, outperforming standalone models. Monte Carlo Dropout incorporated FCN quantify prediction confidence. cost‐effective methodology provides insights into propagation soft viscoelastic media contributes field experimental mechanics. With further data collection, holds implementation, directly linking surface mixture or field‐scale behavior.

Язык: Английский

---

Automated indoor 3D scene reconstruction with decoupled mapping using quadruped robot and LiDAR sensor

Computer-Aided Civil and Infrastructure Engineering, Год журнала: 2025, Номер unknown

Опубликована: Март 4, 2025

Abstract Advancements in automated 3D scene reconstruction are essential for accurately capturing and documenting the current state of buildings infrastructure. Traditional relies on laser scanning to obtain as‐built conditions, but this process is often labor‐intensive time‐consuming. This study introduces an optimization algorithm incorporating methods viewpoint generation, occlusion detection culling, robot‐moving trajectory identification. Additionally, research investigates methods, comparing coupled decoupled approaches identify most practical configuration robotic scanning. Automation strategies collision avoidance human‐centric environments also explored, with adaptive control tested validated efficient point cloud data capture indoor environments. advances state‐of‐the‐art by providing a more precise framework reconstruction. The results demonstrate effectiveness proposed method achieving high scan completeness sufficient density data, offering solutions

Язык: Английский

---

A computational method for real‐time roof defect segmentation in robotic inspection

Computer-Aided Civil and Infrastructure Engineering, Год журнала: 2025, Номер unknown

Опубликована: Апрель 3, 2025

Abstract Roof inspections are crucial but perilous, necessitating safer and more cost‐effective solutions. While robots offer promising solutions to reduce fall risks, robotic vision systems face efficiency limitations due computational constraints scarce specialized data. This study presents real‐time roof defect segmentation network (RRD‐SegNet), a deep learning framework optimized for mobile platforms. The architecture features mobile‐efficient backbone lightweight processing, defect‐specific feature extraction module improved accuracy, regressive detection classification head precise localization. Trained on the multi‐type dataset of 1350 annotated images across six categories, RRD‐SegNet integrates with damage identification tracking. system surpasses state‐of‐the‐art models 85.2% precision 76.8% recall while requiring minimal resources. Field testing confirms its effectiveness F1‐scores 0.720–0.945 types at processing speeds 1.62 ms/frame. work advances automated inspection in civil engineering by enabling efficient, safe, accurate assessments via

Язык: Английский

---

Excavator 3D pose estimation from point cloud with self‐supervised deep learning

Computer-Aided Civil and Infrastructure Engineering, Год журнала: 2025, Номер unknown

Опубликована: Май 3, 2025

Abstract Pose estimation of excavators is a fundamental yet challenging task with significant implications for intelligent construction. Traditional methods based on cameras or sensors are often limited by their ability to perceive spatial structures. To address this, 3D light detection and ranging has emerged as promising paradigm excavator pose estimation. However, these face challenges: (1) accurate annotations labor‐intensive costly, (2) exhibit complex kinematics geometric structures, further complicating In this study, novel framework proposed full‐body directly from point clouds, without relying manual annotations. The parameterized using parameters primitives under kinematic constraints. A unified deep network designed predict clouds. initially pre‐trained synthetic data provide parameter initialization then fine‐tuned real‐world data. facilitate label‐free training, the self‐supervised loss functions exploiting consistency between clouds excavators. Experimental results construction sites demonstrate effectiveness robustness method, achieving an average accuracy 0.26 m. method also exhibits performance across various operational scenarios, highlighting its potential applications.

Язык: Английский

---

Cost‐effective excavator pose reconstruction with physical constraints

Computer-Aided Civil and Infrastructure Engineering, Год журнала: 2025, Номер unknown

Опубликована: Июнь 1, 2025

Abstract Excavator safety and efficiency are crucial for construction progress. Monitoring their 3D poses is vital but often hampered by resource accuracy issues with traditional methods. This paper presents a method to reconstruct the of excavators using cost‐effective monocular camera while considering physical constraints. The approach involves two steps: deep learning identify 2D key points, followed excavator kinematic models, coordinate transformation, projection relationships optimization. Experimental results show achieves mean joint position error 428.58 mm cylinder length 5.12%, outperforming alternative can be employed cost‐effectively monitoring productivity management on sites.

Язык: Английский

---

Transforming Robots into Cobots: A Sustainable Approach to Industrial Automation

Electronics, Год журнала: 2025, Номер 14(11), С. 2275 - 2275

Опубликована: Июнь 3, 2025

The growing need for sustainable and flexible automation solutions has led to the exploration of transforming traditional industrial robots into collaborative (cobots). This paper presents a framework conversion conventional safe, intelligent, cobots, leveraging advancements in artificial intelligence computer vision principles circular economy. proposed modular contains key components such as visual perception, cognitive adaptability, safe human–robot interactions, reinforcement learning-based decision-making. Our methodology includes comprehensive analysis safety standards (e.g., ISO/TS 15066), robot typologies suitable retrofitting, sustainability strategies, including remanufacturing lifecycle extension. A multi-phase implementation approach is laid out theoretical design contribute development cost-effective environmentally responsible robotic systems, offering scalable solution extending usability social acceptance legacy platforms settings.

Язык: Английский

---

Industry perception of competencies for human—robot collaboration in the construction industry: A Delphi study

Frontiers of Engineering Management, Год журнала: 2025, Номер unknown

Опубликована: Июнь 5, 2025

Язык: Английский

---