Self-Adapting Software for Cyberphysical Systems
The built environment — the buildings, utilities, cities and other constructed elements of theanthropocene — is becoming increasingly digitized. The complex array of equipment, sensors and other devices in these environments constitute cyberphysical systems which produce an incredible volume of data. However, this cyberphysical data is hard to access and understand because of the extreme heterogeneity and scale of the built environment: essentially every cyberphysical system is a custom-built “one-off” collection of equipment, devices and data sources that has been continually operated, retrofitted, expanded and maintained over years or even decades.
This dissertation argues that existing barriers to widespread adoption of software-driven sus-tainable practices can in part be overcome through the adoption of rich, semantic metadata which enables the mass-customization of data-driven cyberphysical software. Applications will be able to query their environment for the contextual clues and metadata that they need to customize their own behavior and discover relevant data.
To realize this vision, this thesis proposes a linked-data ontology — Brick — which formallydefines a graph-based data model for describing heterogeneous cyberphysical systems, and a set of ontology design principles for generalizing Brick to other domains. Brick models are created and maintained through a continuous metadata integration process also developed in the dissertation. New programming models are introduced which use graph-based metadata to implement self-adapting applications. Lastly, the thesis develops a novel data manage- ment platform, Mortar, which supports storing, serving and managing semantic metadata at scale. This demonstrates that standardized metadata representations of cyberphysical environments enable a fundamentally richer set of data-driven applications that are easier to write, deploy and measure at scale.