A formalism and a framework for planning of sensor-based construction inspections

Current construction inspection planning practice is not adequately formalized to support the consideration and comparison of multiple possible inspection plans. This leads to the generation of sub-optimal, and possibly inapplicable, plans. This ineffective result is more pronounced in view of the increasing number of technologies available for data collection, and the possibility afforded by a number of sensing technologies to address multiple inspection goals simultaneously. A formal approach for inspection planning is needed to improve the quality and number of feasible plans for given inspection goals, and to search among the feasible space for inspection plans to implement on site. In this thesis, I develop the requirements for representation and reasoning to support automated construction inspection planning; present an automated inspection planning approach that addresses these requirements to support automatic generation of detailed inspection goals and inspection plans; and present results of investigations in reasoning to explore a large inspection plan space resulting from implementation of this approach. First, I present the requirements for representation and reasoning for construction inspection planning based on observations on case studies, documented construction inspection examples, and documented issues related to inspection planning in construction inspection literature. I validate these requirements by investigating the feasibility of construction-site application of these requirements and by identifying the impact of implementing these requirements compared to the status quo as observed on construction sites. Second, I present the approach that I have developed to address these requirements, providing details of the representation and reasoning needed to develop goals and plans for inspection. I validate this work by investigating whether the approach can address inspection contexts found in construction inspection literature. Finally, I discuss my investigation of exploration of the search space of feasible inspection plans that are generated using the inspection planning approach. As this is a component-based approach that enables consideration of multiple technologies and methods, the search space for applicable inspection plans can be quite large. I characterize the size and ruggedness of search spaces that are generated, demonstrate multiple reasoning mechanisms (e.g., heuristics, such as standardizing solutions, and search algorithms, including genetic algorithm, simulated annealing, and random search) that I have investigated for their ability to explore the space. I validate this search-based approach to inspection planning by presenting case studies that demonstrate the difference that this approach makes in comparison to traditional plan selection approaches. This body of work provides three main contributions: a set of requirements and a process concept for developing and reducing construction inspection plan spaces; a component-based approach for development of inspection plans for the construction domain; and a formulation of inspection plan space search that can support construction planning.