Domain modeling of constructed facilities for robotic applications

The introduction of robotics to construction represents one possible solution to some of the productivity and quality control problems facing the construction industry. Construction robots performing operations and maintenance functions in a constructed facility must possess knowledge about and a description of the facility in which they operate. This information about the facility is organized into a domain model of the operating environment. The controller of a robot uses the domain model to plan and carry out robotic actions. This work identifies and investigates issues in the domain modeling of constructed facilities for robotic applications by developing and evaluating a domain modeling system. Model requirements regarding model contents and functionality are first identified and used to develop a model architecture. The domain model must contain both descriptive and functional information about the facility it represents; it must also provide multiple abstract views of the environment in support of planning activities. The architecture organizes the model as a hierarchy of objects obtained through a spatial decomposition of the domain, and uses a network representation to model decomposed objects. Nodes in the network can be either primitive or domain objects. Primitive objects represent non-decomposable physical entities. Domain objects represent aggregate objects and can be recursively decomposed into subobjects: either domain or primitive objects. Connection objects are the branches in the network decomposition of a domain object and represent the functionality of relationships between the subobjects of a domain object. Furthermore, the model architecture uses an object-centered approach to organize and integrate all the information about a domain entity in the model object representing the entity. Modeling situations are used to investigate the limits of the modeling capabilities of the architecture and to identify additional issues that were not addressed when the architecture was developed. Solutions to these issues are generated and their impact on the model structure is used to evaluate the model architecture and to recommend extensions of this work.