Back to search

Clemson University

Reinforcement Learning Based Design Methodology for Building Performance: A Case of Building Facades with Kinetic Elements

Abstract

dc:description.abstract

With the increasing complexity of design problems in building performance, traditional design methods are difficult to meet the growing demand of designers. For example, in building facades with kinetic elements, traditional design methods are facing many constraints due to the complex design variables and requirements. This study applied reinforcement learning(RL) to building performance optimization and proposed a novel design methodology. This design methodology consists of two parts: (1) a reinforcement learning-based design system; (2) an improved design process based on reinforcement learning. The construction of the design methodology started from the analysis of design algorithms. Next, the design system was built based on Python and Grasshopper. Then, this research proposed a design process that incorporates the reinforcement learning algorithm. Finally, a full factorial experiment was conducted to verify the generalization and effectiveness of this design method in different scenarios. Results of the experiment showed that kinetic facades generated by the novel design method perform better than facades generated by traditional design methods in terms of blocking radiation heat and glare. The application of reinforcement learning in architecture is still in the exploratory stage and has many unexplored research directions. By proposing a feasible and efficient reinforcement learning-based design methodology, this study will improve the performance of buildings and provide references for applying reinforcement learning in design.

Degree

thesis:*
Name thesis:degree_name
Doctor of Philosophy (PhD)
Level thesis:degree_level
Dissertation
Discipline thesis:degree_discipline
Planning, Design, and the Built Environment

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Dai, Sida
Contributors dc:contributor
  • Michael Carlos Kleiss, PhD
  • Winifred Elysse Newman, PhD
  • Joseph Choma, PhD
  • Brandon Ross, PhD

Subjects

dc:subject × 6

Identifiers

dc:identifier.*
Repository record dc:identifier
https://open.clemson.edu/all_dissertations/2928

Chain of custody

source
Harvested from
Clemson University
Base URL
open.clemson.edu/do/oai/
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
citation

Dai, Sida. Reinforcement Learning Based Design Methodology for Building Performance: A Case of Building Facades with Kinetic Elements. Dissertation thesis, https://open.clemson.edu/all_dissertations/2928