How to build Genetic Models
April 12, 2012 in Compuational Design by Sivam
An introduction to genetic modelling. Instead of jumping into complex 3D shapes, genetic modelling is explained in 2D. Once the core principles are understood, they can be applied to more complex designs.
Play with 2D genetic models in our learning page >
Feel free to comment. You can download this document if you wish.
Hi Sivam,
A) I’m not very clear with the concept of ‘capturing commanality’ in relation to Genetic modelling. Do you mean to say:
i-never to start the design with pre-concieved ideas? or
ii – Always start the complex model with simple components at the start and than build-it up so that it is ready to explore the outer space? for example (not a genetic example) , if i want to sculpt a complex form of an aeroplane i start with a cuboid slab abd sclupt my way through to build an aeroplane?? is that the logic behind ‘capturing commanality’
B) Concieving geometry in Layers – Do you mean modelling with ‘explicit history’ wher every new component / parameter is overlaid over the previous…in tht way the geometry is layered like in Grasshoppper??
C) isn’t Highly variable model directly proportions to number of parameters?
D) What is the difference between ‘Designer driven generative design process & automated optimisation process’? I could google and do my research on this, but i want your point of view.
Good questions Agneesh.
A) I’m not very clear with the concept of ‘capturing commanality’ in relation to Genetic modelling. Do you mean to say:
i-never to start the design with pre-concieved ideas? or
> All your preconceived ideas are not new. They are new only at one level. Al a level bellow that you will discover too much commonality if your look for it. You need to capture that.
ii – Always start the complex model with simple components at the start and than build-it up so that it is ready to explore the outer space? for example (not a genetic example) , if i want to sculpt a complex form of an aeroplane i start with a cuboid slab abd sclupt my way through to build an aeroplane?? is that the logic behind ‘capturing commonality’
>No, its about capturing what is geometrically common in all aircraft ; fuselage, wings tail etc. Take a look at this : http://www.flickr.com/photos/genoform/4088872905/in/set-72157622643020121
B) Concieving geometry in Layers – Do you mean modelling with ‘explicit history’ wher every new component / parameter is overlaid over the previous…in tht way the geometry is layered like in Grasshoppper??
>Yes, explicit history.
C) isn’t Highly variable model directly proportions to number of parameters?
>Inversely. You will discover that if you play with the lesson in the learning page. In a single layer model more parameters will create noise. If it is layered it will create complexity – of a desirable kind.
D) What is the difference between ‘Designer driven generative design process & automated optimisation process’? I could google and do my research on this, but i want your point of view.
>Both use genetic models, but a designer driven process is based on an open ended search that the designer directs whereas optimization is target driven. You need to set a target. More on this in the following lecture notes.
Thanks sivam,
with regards to ‘capturing commnality’…i have been reading that again and again to understand the underlining meaning. And now i feel stupid to have asked that question…all you mean is to explore the design space at a level where we start innovating than creating what already has been done b4…and with complex geometries and patterns around we should train ourselves to differentiate what is common and what is novel.
I played with the lesson in learning page.
There are only 2 parameters and the range was 50 points which created so many variations.
However, these parameters and range was for low frequency geometry, how did you create and superimpose high frequency on it? Because, high frequency geomtries seemed to have more than 3 points / vertexes as set in original parameters.
Nothing to feel stupid about. We are covering new ground here. No one really knows how to create genetic models. These are guidelines.Once you start taking a genetic perspective the way you see designs will change. You will develop a sense of what is common in families of design and that is the most important skill that you need to develop. Its a visual perceptive skill. Representing it in CAD is relatively easy.
In the demo each side is broken up into n number of points and a random value is added to the points to create the higher frequency form. You can do the same in CAD by creating a base geometry and constructing a higher frequency geometries on top of the base. What we have online is a conceptual model of how it can be done.
You will learn how to do that in the next exercise in the generative design group.