An approach for fusing data from multiple sources to support construction productivity analyses

Construction companies need to continuously improve construction productivity performance to stay competitive. Frequent productivity monitoring of on-going construction activities helps in assessing a project's performance and enables identification of opportunities for improvement. Such productivity monitoring requires data to be fused from multiple data sources since a single data source provides only a portion of the data necessary to perform a comprehensive productivity analysis.;The advent of reality capture technologies, such as Global Positioning System (GPS) devices, Radio Frequency Identification Tags (RFID), laser scanners, onboard instrumentations (OBI) installed in construction equipment, along with the existing project-specific data sources, such as time-card, schedule and as-designed drawings, provides opportunities for capturing and recording what is happening at a job site. To leverage the available data sources to support construction productivity monitoring, there is a need for a formal and generalized approach for data fusion since existing data fusion approaches in construction management domain are designed to work for a specific task, such as defect detection and tracking of materials at a job-site, and do not necessarily work for tasks that they were not designed for.;The main objective of my PhD research is to develop a formalized multi-source data fusion approach to support construction productivity monitoring. In developing such a formalized approach, I have tried to answer two main research questions. The first research question aims at formalizing the generation of a customized process based on user queries for fusing multiple data sources to support construction productivity analysis. The second research question aims at developing temporal and spatial reasoning mechanisms to support the generation of fused data to address productivity related queries of construction engineers and managers.;In order to develop a customized process for fusing multiple data sources, I have tried to answer three main sub-questions. The first question aims at developing an approach to capture user queries related to productivity monitoring. I developed a declarative language to capture user queries so that a project engineer can express his/her queries in natural language and yet it is computer interpretable, and hence it would be possible to automatically identify data items and their levels of details required in a query. The second sub-question is to identify a set of applicable data sources from a set of available data sources to answer a given user query. In addressing this second sub-question, I have developed a data fusion ontology which captures the following characteristics of a data source: (a) level of detail, (b) representation, (c) reference system and (d) list of data items. I also developed a reasoning mechanism based on graph theory that utilizes the ontology to identify applicable set of data sources.;Once the applicable set of data sources are identified, the next problem is to identify a sequence of steps to generate fused data from the applicable set of data sources. I developed two approaches, based on the GraphPlan and Hierarchical Task Network (HTN) specifically, to generate customized sequence of steps (i.e., data fusion plan) to generate fused data. The second research question is aimed at developing temporal and spatial reasoning mechanisms to synchronize the varying levels of details of data sources to enable the generation of fused data for construction productivity monitoring. I developed two different approaches: (a) interpolation and (b) closest neighbor to deal with synchronization problem of spatial and temporal data sources during merging.;To validate the generality of the query capture language and the developed ontology, I used different types of productivity-related user queries, different types of construction-related data sources having different levels of details, representations, and data items. The developed spatial and temporal reasoning mechanisms are validated based on the accuracy of two different approaches. The main contributions of the my PhD research are: (a) query capture language, (b) data fusion ontology, (c) reasoning mechanism to identify applicable data sources, (d) reasoning mechanisms to generate data fusion plan, and (e) spatial and temporal reasoning mechanisms to generate fused data. The practical implication of my PhD research is that it can possibly assist in leveraging different underutilized and valuable data sources for construction productivity analyses.