Project Gallery
“[Early-stage design tools] should enable fluent interaction between designers and knowledge resources in a way that does not hinder the conceptual design process.”
Candy and Edmonds – Design Studies
This thesis identifies and addresses key challenges with the usability of Performance Driven Design (PDD) tools during the early design stages, specifically focusing on the necessary negotiation between quantitative and qualitative design goals. Many aspects of design cannot be computed requiring workflows which integrate the strengths of both architect and algorithm. While algorithms excel at evaluating and optimizing quantitative criteria such as structure or energy, many design decisions rely on the architect’s intuition and expertise for qualitative aspects such as program and circulation. Delegating too much control to algorithms diminishes architect agency while underutilizing algorithms can fail to meet performance targets.
Existing PDD tools, due to their setup complexity, long run times and overly quantitative focus disrupt the architect's dynamic, fast paced and exploratory process. A floor plan design tool prototype demonstrates a comprehensive approach to performance driven design which integrates with the architect's existing CAD-based workflow facilitating the negotiation of space and structure. The tool showcases methods for real time structural evaluation via surrogate models, powering interactive-time design space sampling and optimizations. Backpropagation and differentiable programming techniques enable the use of highly efficient gradient-based optimization algorithms. These optimizations and sampling routines are combined with interactive visualization methods to illustrate and compare the performances of many column positions and the shapes of slab boundaries. The speed and accuracy of prediction and robustness of the optimizations are benchmarked, and the usability of the interactive visualizations is evaluated with a user study.
ITECH M.Sc. Thesis Project 2025: Floorplan Co-Configurator - A performance-driven design tool for balancing space and structure
Lianhan Huang, Jack Otto
Thesis Advisers: Hans Jakob Wagner, Zhenxiang Huang
Thesis Supervisor: Prof. Achim Menges
Second Supervisor: Prof. Thomas Wortmann