Construction performance monitoring has been identified as a key component that leads to the success of a construction project. Real-time and frequent monitoring will enable early detection of potential schedule delays and facilitate the communication of progress information accurately and quickly. To facilitate as-built and as-planned data comparison, this paper proposes an automated registration of a video sequence (i.e., a series of image frames) to an as-planned building information model (BIM) in real time. This method discovers the camera poses of image frames in the BIM coordinate system by performing an augmented monocular simultaneous localization and mapping (SLAM) and perspective detecting and matching between the image frames and their corresponding BIM views. The results demonstrate the effectiveness of real-time registration of images with BIMs. The presented method can potentially fully automate past studies that automate progress inference, given visual representation of as-built models aligned with BIM. Moreover, it will facilitate communication on jobsites by associating quality and progress with visuals that are in the BIM coordinate system.

In the last decade, automated object manipulation for construction applications has received much attention. However, the majority of existing systems are situated in a fixed location. They are mostly static systems surrounded by necessary tools to manipulate objects within their workspace. Mobility is an essential and key challenge for different construction applications, such as material handling and site cleaning. To fill this gap, this paper presents a mobile robotic system capable of vision-based object manipulation for construction applications. This system integrates scene understanding and autonomous navigation with object grasping. To achieve this, two stereo cameras and a robotic arm are mounted on a mobile platform. This integrated system uses a global-to-local control planning strategy to reach the objects of interest (in this study, bricks, wood sticks, and pipes). Then, the scene perception, together with grasp and control planning, enables the system to detect the objects of interest, pick, and place them in a predetermined location depending on the application. The system is implemented and validated in a construction-like environment for pick-and-place activities. The results demonstrate the effectiveness of this fully autonomous system using solely onboard sensing for real-time applications with end-effector positioning accuracy of less than a centimeter.

Deviations from planned schedules in construction projects frequently lead to unexpected financial disadvantages. However, early assessments of delays or accelerations during the construction phase enables the adjustment of succeeding tasks and their dependencies. These assessments involve human resources and as-built information that may not be immediately available. To reduce manual processes of collecting and generating as-built information, this paper presents the development and validation of a new pipeline of localizing a camera to a BIM by detecting and matching perspectives of images and the BIM. The proposed method increases the degree of automation of construction progress monitoring. The novel method registers as-built video data on as-planned BIM data in real-time using computer vision algorithms. The results demonstrate the improved registration errors compared to that of the manual correspondence selection and the potential for automating the initial manual steps of data collection and registration of the past studies that attempted to automate progress detection and quality assessment.

I.A.A.R.C. - International Association for Automation and Robotics in Construction Civil Engineering

An autonomous robot that can monitor a construction site should be able to be can contextually detect its surrounding environment by recognizing objects and making decisions based on its observation. Pixel-wise semantic segmentation in real-time is vital to building an autonomous and mobile robot. However, the learning models’ size and high memory usage associated with real-time segmentation are the main challenges for mobile robotics systems that have limited computing resources. To overcome these challenges, this paper presents an efficient semantic segmentation method named LNSNet (lightweight navigable space segmentation network) that can run on embedded platforms to determine navigable space in real-time. The core of model architecture is a new block based on separable convolution which compresses the parameters of present residual block meanwhile maintaining the accuracy and performance. LNSNet is faster, has fewer parameters and less model size, while provides similar accuracy compared to existing models. A new pixel-level annotated dataset for real-time and mobile navigable space segmentation in construction environments has been constructed for the proposed method. The results demonstrate the effectiveness and efficiency that are necessary for the future development of the autonomous robotics systems.