Due to the frequent occurrence of occupational accidents, the construction industry has been regarded as one of the most dangerous industries, causing a lot of causalities and economic losses every year. Statistics show that collision with heavy construction machines in operation is the main source of on-site hazards. Monitoring the operations of heavy machines on a construction site is crucial to prevent potential collisions and improve operational safety.

Currently, automated monitoring of unsafe proximity and potential collision has been achieved preliminarily by acquiring the 2D location information of the construction machine. However, construction machines, especially articulated machines with multiple independent movable components, have dynamic and complex spatial poses. For exa...[ Read more ]

Due to the frequent occurrence of occupational accidents, the construction industry has been regarded as one of the most dangerous industries, causing a lot of causalities and economic losses every year. Statistics show that collision with heavy construction machines in operation is the main source of on-site hazards. Monitoring the operations of heavy machines on a construction site is crucial to prevent potential collisions and improve operational safety.

Currently, automated monitoring of unsafe proximity and potential collision has been achieved preliminarily by acquiring the 2D location information of the construction machine. However, construction machines, especially articulated machines with multiple independent movable components, have dynamic and complex spatial poses. For example, although the machine stays in a location, the moving independent components are still likely to threaten operational safety by colliding with other construction-related objects in 3D space. Therefore, tracking poses of construction machines is important in preventing potential collisions and unsafe proximity, yet it has been neglected in previous studies.

This research aims to develop approaches to estimate the full-body pose of construction machines based on onboard sensors, and thus assist in improving operational safety on sites. There are two parts involved in the research: (1) Pose estimation of construction machines based on homogeneous sensors; (2) Pose estimation of construction machines based on heterogeneous sensors and application on operational safety.

For pose estimation of construction machines based on homogeneous sensors, the information requirements on pose-related analysis of construction machines using inertial measurement unit (IMU) data are defined. Based on the proposed information flow, the current full-body poses of construction machines can be estimated using a kinematics model and IMU data. Moreover, an optimal installation scheme for IMU sensors is investigated systematically to optimize the location and number of IMUs to install.

For pose estimation of construction machines based on heterogeneous sensors, first, a full-body pose estimation framework is proposed for excavators, the typical articulated construction machine, with a loosely coupled data fusion strategy to utilize different types of onboard sensors for enhanced accuracy and robustness. Specifically, a non-invasive onboard visual-inertial sensor system is designed for data fusion. Then, through loosely coupled competitive and complementary data fusion, the keypoints describing the full-body poses of the excavator are tracked in 3D space. Especially, an EKF-based localization algorithm is developed for optimized multi-keypoint tracking, which is verified to improve the accuracy and robustness of pose estimation by a real-world excavator case study. In addition, to apply high-accurate pose estimation to operational safety monitoring, a proximity warning framework based on tightly coupled sensor fusion is designed for excavators to achieve highaccurate and stable collision prevention. In the framework, an optimization-based visualinertial estimator is proposed to obtain the optimal 3D locations of multi-keypoint of an excavator, which is validated by real case studies. Especially, the kinematic and structural constraints of the excavator are imported into the developed estimator to enhance the accuracy of estimation.

Compared to current practices, the proposed research provides a theoretical basis for economical and effective full-body pose estimation of construction machines using onboard sensors and developing an accurate and robust 3D dynamic proximity warning system for improving operational safety on construction sites. It is expected that the proposed research will help to achieve more precise and detailed operational safety management on construction sites in the future.