Material Management and Progress Detection for Construction on Earth and Beyond

Open Access
- Author:
- Stephans, Tyler
- Graduate Program:
- Aerospace Engineering
- Degree:
- Master of Science
- Document Type:
- Master Thesis
- Date of Defense:
- November 03, 2023
- Committee Members:
- Amy Pritchett, Program Head/Chair
Alan Richard Wagner, Thesis Advisor/Co-Advisor
Robert G. Melton, Committee Member
Rob Leicht, Committee Member - Keywords:
- mars
construction
automated construction
robot
simulation
material selection
progress monitoring
material management
robotics
computer vision - Abstract:
- Various construction projects stand to benefit from automating on-site material management. In the construction industry, moving materials between stockpiles, staging areas, and workface locations costs time that could be spent furthering construction progress. Handling materials can also be a safety hazard, and current material management practices often result in supplies being wasted. Further, as NASA and commercial companies seek to build infrastructure on other worlds, the management of materials in these projects will have to be fully automated. This thesis presents an automated process where materials are selected for staging or assembly by detecting construction progress from site images. The material selection system uses geo-referenced images to create a point cloud of the structure being built, and the point cloud is compared with a 3D model of the planned structure for varying levels of progress. Through this comparison, the current state of progress of the structure is determined and used to select materials. The materials may then be delivered to their respective workfaces or used for assembly by a robot. Experiments were initially conducted in simulation using an environment designed to resemble a construction site on Earth. Physical experiments were then performed on a mock construction site. Finally, an experiment was conducted in a Mars simulation environment. All experiments were evaluated with respect to their ability to select the correct materials for staging based on the detected construction progress. Lighting, camera location, occlusions, material shape, and current construction progress were varied to assess the generality of the process. In all tested types of environments and under various conditions, the material selection was successful. Unscheduled occlusions and lighting were found to have the greatest impact on system performance, which led to indirect sensitivity to camera location and material shape. A camera can observe varying levels of lighting depending on the relative positions of the camera, structure, and light source. The material's size and shape can also make it more or less likely that its detected progress is affected by occlusions and alignment error. When occlusions were absent and enough lighting provided, material selection was successful. A mobile robot was then used to demonstrate automated structure assembly on Mars based on successful material selection. The system presented in this thesis contributes to the feasibility of autonomously managing materials on a construction site and reveals potential avenues for future research.