Module 8 Questions:
- How does defining the design objectives influence the outcomes in a parametric design study using Generative Design in Revit? Provide examples of how different objectives might result in varied design options.
Defining the design objectives influences the outcomes in a parametric design study by defining the standards by which we evaluate outcomes. An ideal solution is not objective: we must tell the model what parameter or standard we are optimizing for so that the solver knows exactly what goal to move towards.
For instance, a generative design study that optimizes for cost would likely simplify structural shapes whereas something that optimizes for aesthetics (which is of course much more difficult to quantify) wouldn’t shy from non-structural ornateness.
- Discuss the importance of identifying target taxonomies when generating synthetic datasets in architectural design. How can this help in managing large datasets and ensuring diversity and accuracy in your designs?
Target taxonomies need to be categorically differentiable, algorithmically representable, reproducible and scalable, and result in modular generative workflows. It’s important to make sure that within target taxonomies we avoid overlaps and repetition or bias.
It’s crucial that target taxonomies are identified prior to generating synthetic datasets so that the data is relevant and useful. This can help us by streamlining the variables that we establish so that datasets do not become tremendously large, while also ensuring that the designs account for significant diversity in input parameters. It’s much wiser to spend time selecting proper target taxonomies that will output meaningful results than to generate excessive data that needs to be sorted through and “cleaned” in post-processing.
- What are the potential benefits and challenges of using automated workflows for generating diverse synthetic datasets in parametric design? How can modularity and scalability be achieved in such workflows?
Potential Benefits:
- Diverse synthetic datasets give info
- Saves time by automating data generation
- Expands understanding of design space beyond initial assumptions (i.e. computer might facilitate a more creative design choice by widening designer’s vision of what’s possible within a set of design boundaries)
Potential Challenges:
- Computationally expensive for large systems
- Significantly less existing training data
- Data sets aren’t diverse or complete enough
- Might reinforce design biases
- Explore the role of iterative processes in optimizing design options within the generative design framework. Why might it be necessary to tweak and repeat studies, and how can this approach lead to better design outcomes?
It will never be perfect on the first try! A lot can be learned from the first round of results which will then result in tweaking parameters for further studies. An initial design study might generate outputs that are not physically feasible (i.e. introduce a previously ignored constraint that the user must implement for future studies).
Questions Related to the Autodesk Class:
- Describe the general workflow of creating a generative design study in Revit, as presented in the lecture. What are the key steps involved, and how do they contribute to the generation of optimized design options?
General workflow:
- Establish design variables and objectives.
- Reduce the variables to inputs and the objects to ranks/outputs. This allows the designer to define goals quantitatively. It also provides a mechanism by which to compare outputs, which is crucial since not all design parameters can be weighed objectively. In a multi-objective study, for instance, perhaps there are two evaluation parameters that hold different levels of importance for the designer. This step allows the designer to think carefully about how each input and output is considered and weighed numerically, and to ensure that this bias is identified and selected (and not left to the computer to enact).
- Select appropriate solver, then start generating design options. Selecting an appropriate solver is critical, see responses to question 4.
- exploring these options, and then repeating with tweaked study until happy with the design options outputted. Any good design process is circular!
- Given the properties of a good synthetic dataset outlined in the class, such as being large, labeled, parametric, expandable, diverse, and balanced, how would you apply these principles to create a dataset for a specific building morphology study using Dynamo? Outline your approach.
My approach would be as follows:
- Define my design goals —> what am I optimizing for? Or am I just trying to explore the design space?
- Simplify the design space —> make sure I’m isolating the exact problem I want to address. Start with few input parameters and only 1 or 2 clearly labeled evaluators.
- Automate simulations with initial set of parameters —> are my outputs wide enough? If optimizing, is my algorithm favoring one end of the spectrum of my input parameters? I.e. do I need to widen/adjust the range?
- Study the output quantitatively, and judge which results are meaningful
- Tweak and repeat!
- Identify and discuss the four different solvers mentioned in the lecture that can be used to generate building masses in the Generative Design tool. How do these solvers impact the sample space and variety of design options produced?
- Randomize solver assigns random values to each input
- Cross-product solver generates combinatorial datasets —> gives outputs for all possible input parameter combinations
- Like-this solver takes in a current design and then applies small tweaks in parameters
- Optimize solver iterates multiple designs to optimize for a specific objective (or objectives).
Different solvers should be used for different problems. Like-this solvers are useful in situations with an existing base-solution, like for sensitivity studies. A randomize or cross-product solver will generate a wider array of diverse options, from which you can narrow down to a final design choice. Depending on the solver you use, your sample space and variety of design options will be bigger or smaller, more or less diverse.