Introduction: Generative Design - Evolution of a Digital Bonsai Tree

I started working with the Research Group at Autodesk with Dreamcatcher about 2 years ago.

At that time I was using it to design spacecraft.

Since that time I have learned to love this software tool as it allows me to explore thousands of designs, and options that I would have never imagined due to time constraints.

It also has a feedback loop, I can't un-see a design, I am inspired by many of the designs that emerge along the iteration path.

The path is very much like watching time zoom by. 1000 years per 24hrs. The only thing I have really understood that is 1000 years old is the redwood trees in California, Big Basin in particular. There is a tree that marks on the rings the dates of events in History. It's very humbling to walk among those trees.

Dreamcatcher solutions like trees are a product of the environment and the loads and constraints that are applied during its life time. A little change and the tree could be totally different.

Step 1: Generative Design, Existing Geometry - Ports

First some background.

If you have two blocks, A. and B.

You would like to connect them.

Dreamcatcher will synthesize geometry based on your input.

Step 2: Generative Design, Existing Geometry - Obstacles

If you want your connecting geometry to only be on the faces that are closest to each other, and nowhere else you must MASK the part so that the new synthesized geometry cannot connect.

This will be an additional body or bodies of existing geometry.

Step 3: Generative Design, Initialization - Generate First Iteration

The first iteration includes all existing geometry.

(this can be in the shape of something else using Seed geometry - not covered here ;) )

Step 4: Generative Design, Generate and Evolve

Based on your applied loads and solver settings you will get synthesized geometry connected to your existing Ports A. and B.

It's important to note that the future of tools like this is far beyond making optimum anything, it's not clear that the final form will ever make structures. Right now the tool exists. This is amazing. It can do topology optimization and truss optimization at the moment. It is used to synthesize geometries.

The process of high level constraints or goals, and a computational curation which searches thousands or millions of potential solutions will change our lives in ways we cannot imagine. That is the power of using this tool today. You can get a glimmer of the future and it is RAD!

Lets make a lamp!

Step 5: Generative Design: Early Failure, to Group Project, to Growth of a 1000 of Years!

As one of the first things I did at the Pier I set out to make a lamp.. the first attempt fell flat on its face, but we got an opportunity to do a group project. The lamps evolved from that collaboration.

Step 6: Group Project


We wanted to create a beautiful lantern to bring warmth to our space at Pier 9. (Air-ea)

We utilized two separate design approaches, one for the lantern base, and one for the lantern illumination component.

Prototypes were printed on Makerbot Replicator 2 as well as the Objet Electronics prototypes were sized for the larger version, but due to time constraints were modified to fit a scaled down version.

Design Team:

Charlie Katrycz

Mary Franck

Michael Koehle

Michael Vergalla

Step 7: Start With Existing Geometry

The Top Ring and Bottom Ring are Ports

The Rest are Obstacles

The inputs to Dreamcatcher were designed in Fusion 360

There are two important CAD features that become the loaded port, and the fixed port.

The ring that the lamp hangs from. (Fixed)

The ring that the lighted part of the lamp sits in. (Loaded)

The rest of the geometry are arbitrary obstacles that allow for the bulb to be placed in the ring, and help control some keep out zones such as inside the hanging ring. This is so nothing grows where I want to put a loop of string through to hang the assembly.

After tuning generation parameters the model is "put in the oven"

The next morning I have ~2000 designs based on my constraints.

The first prototypes were generated without the last 3 large lobes. (2nd Image)

The first prototype did not allow for easy integration with the bulb assembly.

Step 8: Design Exploration, Selection, Integration

I ran 3 Problem definitions with different load cases, and solver parameters.

By Monday morning I had 6000 designs!!!!


The fun part about dreamcatcher is that you create interesting creatures that need exploration.

I took a subset of the designs that were converged and some that were on the event horizon and started printing them for the team to discuss.

In the mean time I will chase this alien design as far down the rabbit hole as I can!


Ohhh look how cool it looks!!!!

The egg designs are awesome because we are looking at a variety of ways to play with light through multi-material 3D Printing processes. Complexity is "Free" (i used to hate that cliche)

---------------- Lantern Base Prints are done Time to select as a team!

We have a winner!

First Image, Far Left.

We plan to print one with a height of 170mm in the Z axis, but the Fortis was not available, so we have to make a final egg lantern component that fits the small models (100mm in the Z axis)

Step 9: Final Group Lantern

It all comes together, and its glorious!

Both final egg prints, when lighted give a warm, welcoming glow.

Combined with the dreamcatcher bases we have accomplished what we set out for:

A one of a kind lantern that utilized cutting edge design software and parallel work flows.

Step 10: Digital Bonsai

I like the problem and wanted to chew on it some more. The 2 self standing pieces presented at the show are explorations of not a hanging lamp but when problem is flipped upside down.
I ran lots of cases, with incremental increasing load cases and designed the electronics mounting in Dreamcatcher as well. There are moments where the design takes evolutionary leaps. Its really cool.. it heads a certain direction.. and then something changes drastically and it begins to settle. It's one of the reasons I love exploring with this tool, its like watching life happen. The future vision and development is really amazing. Each version of the software is better then the last.

People see digital object in a finished form and something looks missing, sometimes it's the brush stroke or something else without words. The generative design process is a digital craft, a play between the creator and the computational resource. Depending on the person a TI-89 calculator looks like a calculator or a powerful tool.

This will rapidly come into mainstream use and it will be accepted. I think the power is beyond structural components alone, but multiphysics solutions. Solutions that can create communities that balance with microclimates and natural surroundings. Wasting less energy and heat.

The finished part is important. I have struggled with that for the last 6 months. Probably longer! hahaha

3d prints alone at the moment dont connect with people. In some cases there are fashion pieces or enabling designs but a 3d printed object alone does not carry the allure it once did.

How to evolve?

Step 11: Redwood Base

Early in the residency I saw some chunks of "scrap" redwood. I felt connected to the trees and their age impacted me. They were the perfect platform for this work.

At the time of the show I had most of all the electronics ready, but the batteries only lasted a couple hours. I really messed up my timing with the sun. My main work was backlit when I imagined it at night. This work was not lit.. anyway it was a good lesson.

I finished all the electronics units and tested them after the show.

Step 12: Bonsai Tree Bears Fruit

In unconventional form these trees fruit.

The lights are controlled by a phone app that controls color and intensity.

These will look really good in my original vision of them in the woods, in nature. Hanging along a trail.

Step 13: Generative Design, Electronics Interface

I had one experiment running this whole time. I wanted to make the smallest features possible by topology optimization algorithms. The goal of this was to draw a parallel to biomedical components.

Each time I found a solution. I would throw it back into the system.

After a long time I was able to raise the resolution, and keep driving the system to make smaller features.

I was able to print the part on the Autodesk Ember printer. It fit in the diagonal of the print tray.

Step 14: Generative Design, Human Touch

Generative design exists in a number of forms, and to date they are growing in maturity. Very soon it will be everywhere, but in the meantime sexy, odd looking 3d prints, just don't make the cut. Another artist at Autodesk Arthur made a beautiful wooden chair. Generative Chair

I had many conversations asking people who interacted with the pieces what they connected with or didn't.
I was unable to communicate easily, my experience of the process.

The curation with computational assistance.

You can feel the future.
It's so hard to describe or communicate. So at least for me, it gets lost.

For now.

Data processing tools that I have seen glimpses of will help tell the story, the journey of generative design.

For me. I needed to add the human element. I wanted a smooth finish, and I was pointed to wax.

I wanted a strong design that could be handled. It was originally designed in aluminum, but printed in plastic.

Touch was important. For trees, for humans, for art. I also think for generative design!

I took the piece to the Artworks Foundry in Berkeley, California.

After speaking to them they helped me put some wax layers on the part.

Then set me free to work the wax to my satisfaction.

Finally, the HUMAN ELEMENT. It was truly a design collaboration. I think this is part of why the Wooden Chair is so beautiful. You know that there is a craftsman. So maybe that's it. More of the ping pong, a back and forth in different stages of the process.

Step 15: Generative Design, Bronze Casting

Images of part right after casting in bronze.

Vents and pour locations have not been removed!

More to come. Check back later for the next evolution of the Digital Bonsai Tree.