Incorporating Robotic Constructability in Computational Design Optimization
Restricted (Penn State Only)
- Author:
- Zargar, Seyed Hossein
- Graduate Program:
- Architectural Engineering
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- June 06, 2024
- Committee Members:
- Rob Leicht, Major Field Member
Nathan Brown, Chair & Dissertation Advisor
Jose Pinto Duarte, Outside Field Member
Alan Wagner, Outside Unit Member
Caitlin Mueller, Special Member
Julian Wang, Professor in Charge/Director of Graduate Studies - Keywords:
- Autonomous construction
Robotic constructability assessment
Multi-objective optimization
Design space exploration
Data-driven design
Surrogate modelling - Abstract:
- Computational design and multi-objective optimization can empower building designers to improve design quality through multidisciplinary performance simulation. Simulations predicting building behavior in the domains of structure, energy, or daylight can help designers prioritize performance goals and make informed decisions. However, ease of construction remains a crucial aspect of design, as major technological advancements have shifted the AEC industry's focus from ‘can we build it?’ to ‘how should we build it?’. Early design exploration and optimization have the potential to contribute toward flexible construction-related suggestions and guidance, as constructability has gained significant importance due to the potentially transformative uses of robotic technologies. Quantifying constructability is advantageous for early-stage design as it can enable designers to balance performance objectives with practical construction issues. While it is difficult to quantify all factors that influence construction, the introduction of robotic construction processes and their associated simulation sequences offer rich datasets by which to compare potential outcomes. While optimization for robotic construction presents promising advancements, navigating the transition to full-scale building construction, especially with the integration of traditional and hybrid techniques, remains a challenging and open research area. This dissertation examines the inclusion of constructability metrics in early computational design exploration by analyzing how structures should be built with robots for several design scenarios. It contains four interconnected studies: (1) performing a literature review to categorize and assess robotic construction tasks based on their potential use as objective functions or constraints for early design optimization, (2) demonstrating simplified construction-aware design exploration and pinpointing its quantifiable benefits on timber structures with linear elements, (3) developing a new method to evaluate ease of constructability based on assembly sequences of a real-world mobile robot setup during optimization of panelized timber structures, and (4) training and testing machine learning models to predict different constructability metrics associated with robot movement and path planning for new panelized timber structures based on underlying features. Overall, this research aims to explore the positive benefits of incorporating robotic constructability knowledge into the early-stage design workflow and evaluate ways in which this information can contribute to adaptable construction-related suggestions and guidance.