Unmanned ground vehicle-driven indoor monitoring: Enhancing routine data acquisition and on-site deployment

Monitoring is crucial in the construction industry, acting as an essential tool for managing contractors' payment requests based on completed work. However, existing monitoring methods, including paper-based documentation and digital tablets, fall short in fulfilling the demands for routine tracking. These methods face challenges like irregular and infrequent reporting, leading to delays in obtaining accurate performance data. Additionally, the reliability of collected data is often compromised by the subjective judgment of practitioners. Furthermore, the process of data acquisition is marked by inefficiency, demanding considerable time and effort for both on-site data collection and planning the subsequent documentation. Motivated by these identified challenges, this dissertation is dedicated to the design, development, and validation of a novel Unmanned Ground Vehicle (UGV)-based system. The goal is to minimize human involvement in routine data collection and monitoring tasks within indoor settings. For this purpose, a UGV prototype equipped with a depth camera and LiDAR has been developed. This prototype serves as a functional platform for autonomous data acquisition and monitoring activities.During the deployment of the UGV for routine data acquisition and monitoring, an up-to-date navigation map is typically required to enable effective planning, navigation, and localization. This study introduces a video-based pre-mapping method leveraging prior construction progress video data. It addresses inconsistencies caused by dynamic changes of construction sites and produces an updated map, thereby improving the map's utility for planning and navigation. By minimizing the need for labor-intensive on-site map creation, this pre-mapping method significantly enhances the efficiency of deploying the UGV for routine data collection and monitoring tasks.Upon having the navigation map, the developed camera-equipped UGV is programmed to reach waypoints and execute its tasks. However, inherent errors in UGV's sensors and actuators may still prevent the UGV from accurately reaching the designated pose. Considering this, a novel sequential pose rectification method is proposed to correct the UGV camera pose, aiming to align the UGV's perspective closely with the Building Information Model (BIM). Such an approach is expected to enhance the reliability of inferred results, particularly when making comparisons with corresponding areas in BIM.Additionally, once the UGV reaches the waypoints and conducts the data acquisition, it's crucial to automatically detect the presence of elements within this data, typically represented as point clouds. By using material features identified from images, the segmentation of objects within point clouds is improved. This not only increases the efficiency for scene understanding but also boosts the Scan-to-BIM applications. Furthermore, automating the planning for the UGV's data collection mission is crucial for routine monitoring. This study introduces an optimization approach leveraging a Genetic Algorithm to strategically plan and prioritize data collection missions for the UGV, focusing specifically on areas with active construction activities. This consideration enables effective data collection on crucial construction elements while ensuring the acquisition of comprehensive and-quality data.In summary, this dissertation introduces a UGV prototype tailored for data acquisition and indoor monitoring, addressing several challenges encountered with its deployment. Validated in indoor sites, this prototype shows promise for effective operation in nearly completed indoor environments, such as during the handover phase. The insights derived from this study contribute to the further development and enhance the applicability of UGV-based systems in indoor data acquisition and monitoring tasks.