Rethinking Early-Stage Construction Planning Using Virtual Reality and Point Cloud Processing

Abstract

Construction planning is inherently complex, requiring practitioners to define scope, determine feasible execution sequences, and evaluate cost and resource implications. These challenges are amplified in early-stage planning, where decisions made under high uncertainty can significantly impact project performance, particularly in complex industrial facilities such as nuclear power plants, where access is limited and legacy assets often lack reliable drawings or models. Despite this, existing approaches such as Building Information Modeling (BIM) and digital twins rely on time-consuming, high-fidelity modeling processes that are not well aligned with the need for rapid and flexible decision-making for early-stage planning. As a result, there is a need for fit-for-purpose planning approaches that prioritize speed, usability, and decision support over detailed model reconstruction. This thesis investigates two fit-for-purpose uses of digital technologies for early-stage construction planning: virtual reality (VR) as an interactive environment for exploring and generating construction sequences, and point cloud processing as a direct source of geometric information for quantity estimation. Compared to traditional planning tools, VR enhances spatial understanding through intuitive interaction, real-time physics-informed feedback, and realistic simulation enabled by advances in physics engines and game development platforms. This research extends VR beyond visualization by using it as a process generation tool, where user interactions are captured and structured to enable the automatic generation of construction schedules. While VR supports process generation, it does not inherently provide quantitative information about the physical scope of work. To address this, point clouds obtained from laser scanning are leveraged directly for quantity estimation, avoiding labor-intensive Scan-to-BIM workflows. By structuring point cloud data into meaningful objects and applying geometry-based analysis techniques, key quantities such as volumes, areas, and lengths can be computed efficiently. The results demonstrate that these approaches can support faster and more flexible planning processes by reducing reliance on detailed modeling and enabling more direct interaction with planning data. This thesis contributes a fit-for-purpose perspective to construction planning, highlighting how different digital representations can be strategically applied to improve early-stage decision-making.