Motion planning and computer vision for facilitating smart module transportation in modular integrated construction

Abstract

Modular integrated construction (MiC) transforms fragmented site-based construction into integrated value-driven production, with ready-to-use modules being manufactured offsite and transported to sites for assembly, providing the opportunity to enhance the quality, productivity, safety, and sustainability of construction. Despite the many advantages of the modular approach over traditional methods, the module transportation process presents challenges, such as module passing ability analysis, optimal route selection, and strict just-in-time transportation requirements. Moreover, these challenges are amplified in high-rise, high-density cities such as Hong Kong and London. As MiC is a new construction method, little research has been performed for resolving these transportation-related issues, limiting its adoption.

This research aims to investigate the mechanism of the module transportation process, identify gaps between the current research progress and contemporary practices, and develop novel motion planning- and computer vision-based methods for improving the transportation safety, efficiency, and productivity in MiC. This research involved a systematic background study with site visits and observations, the development and validation of four novel algorithms that solve critical tasks related to module transportation, and the development of an integrated artificial intelligence and Internet of Things (AIoT) framework facilitating smart module transportation in MiC. The background study provides the context for the overall research, including the motivation, progress, current practices, and technologies with potential for enhancing the module-transportation process. The first algorithm—Truck Parallel Hybrid A* (TP-Hybrid A*)—is for quantifying the module passing ability in critical transportation scenarios, such as wide turns and narrow roads. The second algorithm—customized routing for module transportation (CR4MT)—allows optimal route selection for module transportation. The third algorithm is a dynamic coordination method for vehicle fleets to achieve just-in-time module transportation. The last algorithm is a computer vision-based module detection model that provides real-time module installation progress information to support the just-in-time module transportation. These four algorithms were validated using data collected from simulations and real-life projects. The proposed algorithms outperformed the benchmarks with regard to accuracy and efficiency. The integrated AIoT framework can connect the algorithms to a system that facilitates smart module transportation in MiC.

This study makes original contributions to the field of module transportation. Tasks in module transportation were formulated as motion planning problems, which promotes an in-depth understanding of module transportation and provides a new perspective for investigating module transportation problems. Methodologically, motion planning methods and computer-vision techniques were extended in the development of the proposed algorithms. Practically, the proposed algorithms and system were validated and demonstrated using case studies, which will help MiC practitioners to implement them in practice. The proposed approaches and corresponding findings can improve the safety, efficiency, and productivity of module transportation. This research should facilitate smart module transportation and the wider adoption of MiC, promoting automation in the construction industry.