Reasoning About Interior Building Design, Grounded on Design Rules

Abstract

Computers have emerged as an invaluable tool in exploring building interior configurations, before committing to a particular layout. Building Information Modeling (BIM) enables designers to create digital representations of alternative interior arrangements and supports computer-automated evaluation of the arrangements under consideration against a variety of construction, accessibility, and stylistic guidelines. This process is known as Automated Code Checking (ACC). Domain Specific Languages (DSLs) can help domain experts code ACC rules in a machine-readable format, focusing only on domain concepts instead of programming knowledge. This thesis provides a coherent suite of algorithms for reasoning about building designs based on RuleDSL, a user-friendly DSL for describing building spaces, their contents, and the geometric relations among them. The thesis puts forward algorithms for (i) automatically generating alternative interior layouts, (ii) evaluating them against a variety of quality metrics, and (iii) automatically learning RuleDSL rules from example layouts. RuleDSL and the above algorithms have been developed and evaluated in BIM-kit, a state-of-the-art software platform that implements and validates the above algorithms as well as supporting experimentation with a variety of use cases in the broad area of design automation.