Generative Design Finding Layout Shapes [ㄱ, ㄴ, ㄷ, ㅁ]

test file Download link

 

Concept for a Form Composition Method for Layout Automation

Finding Layout Shapes [ㄱ, ㄴ, ㄷ, ㅁ]

I conduct experiments to identify shapes within a site that adheres to basic architectural regulations. From the applied building massing, I extract forms of [ㄱ, ㄴ, ㄷ, ㅁ] and implement a layout algorithm capable of generating either a single building block or multiple building blocks.

• By arranging three rectangular masses, one or more building blocks can be formed.
• The three rectangular masses are placed randomly, creating two to three segmented building blocks.
• Depending on the size and orientation of these rectangles, shapes [ㄱ, ㄴ, ㄷ, ㅁ] are produced.

 

Preparation for Layout Automation (Building Regulation Related)

Securing a Buildable Site Area After Applying Regulations

Identify the corner intersection of roads, then find the closest corner point of the site boundary to that intersection. From that point, create a circle with a 2-meter radius and extract the two intersection points where the circle overlaps the site boundary. These points represent a 2-meter setback on each side. The resulting setback zone is then excluded from the building placement area.

• Extract the road intersection.
• Identify the corner point of the site boundary closest to the intersection.
• Create a 2m-radius circle and find the intersection points with the site boundary.
• Connect the two points to form the setback zone.
• Exclude the created setback zone from the buildable area.

 

 

Dynamo Shape Configuration for Applying True North Daylight Regulations

For any building height exceeding 9 meters, the building must be set back at least h/2 from the site boundary line on the true north side.

• This requirement can be implemented in a Revit family as shown in the image below.
• By using a 2-point variable family and applying formulas to the profile, it is possible to create parameters that enforce this setback.

When the above formulas and parameters are visualized, the shape can be formed as shown in the image below.

If this shape overlaps with the building’s form, it indicates a discrepancy from the legally permitted scale and should be treated as an error.

Layout Method for Environmental Performance Evaluation

Setting the South-Facing Ratio and Environmental Performance Values

Use the south-facing ratio ((sum of south + north lengths) / (east + west lengths)) as the baseline formula to measure the length of the building shape. Then, assign this measured length as an output value in Generative Design, and apply filtering operations based on it. As shown in the image below, various forms of environmental performance values can be extracted.

Identify horizontal layout tendencies when the environmental performance value ranges between 1.5 and 3.   

When the environmental performance value is 1.5 or less, a tendency toward vertical layout can be observed.As the number of masses increases, the meaningfulness of directional result values decreases.

Method of Determining Evaluation Criteria (Building Coverage Ratio, Number of Buildings, Building Width, Building Perimeter, etc.)

Check the building coverage ratio, number of buildings, building width (m²), and total building perimeter (m).

Conduct experiments and confirm rules related to form finalization using factors such as building coverage ratio, number of buildings, building width, and total building perimeter.

 

Among 1,000 random outcomes, only 4 were identified as courtyard-type forms.

• There is a need to review an algorithm that can intentionally extract the "ㅁ"-shaped arrangement.

• When generating results, a step must be included to extract those with empty internal spaces (courtyards) inside the solid.

 

Image-Based Shape Recognition

Recognize Image Files and Compare with Dynamo Geometry for Placement

In Dynamo, images can be recognized to generate solid forms.

• Place a ㄱ-shaped solid at the center of the site.
• Retain the black-colored shape area.
• Check the overlapping spaces with the four quadrilateral shapes already

 

 

Setting the [Shape Similarity] Output Using Generative Design

• Measure the volume of the overlapping area between the image shape and the rectangular arrangement.
• The higher the overlapping volume, the more similar the arrangement is considered to be to the image shape.
• However, it was observed that most arrangements do not resemble the image shape.

Setting the [Shape Similarity min] and [Shape Similarity max] Outputs

max: The volume overlap value between the image shape and the layout shape

• A larger number indicates greater similarity in shape.

min: The volume outside the image shape (excluding the overlap)

• A smaller number indicates greater similarity in shape.

The shape with the highest max value (20 points) does indeed have a ㄱ form, but its min value also exceeds 20. This indicates that while the shape resembles the image, it actually creates a broader ㄱ configuration.

For this reason, new ideas and methods are necessary to enhance the correlation between the image shape and the resulting layout.

 

 

 

 

Adjusting the Rectangular Layout Area to Match the Image Shape

Exclude the possibility of placing shapes in locations unrelated to the original form.

Reduce the number of layout permutations, thereby increasing the likelihood of arrangements that conform to the intended image shape.

 

Checking the [Shape Similarity min], [Shape Similarity max] Output Values (Below 10, Above 15)

Increasing the probability of shape placement has resulted in more recognizable forms than before.

However, the arrangement still falls short of achieving a perfect shape.

Consider additional methods to exclude areas that deviate from the intended

 

Recognition and Outcome Generation Through Generative Design Execution

Extract Generative Design results by utilizing the Shape Similarity Min/Max values.

Setting the [Shape Similarity min] and [Shape Similarity max] Output Values (Under 10, Over 15)

Optimization in progress

Results show better shape retention compared to previous outcomes

However, achieving perfect shape similarity still appears challenging.

 

Setting the [Shape Similarity min] and [Shape Similarity max] Output Values (Under 10, Over 15)

Conducting optimization experiments for a ㅁ (courtyard) form

To achieve a shape with an inner courtyard, modifications to the rectangular shape are necessary

test file Download link