0:00 This crowd. End of the day. Is it like Is it working? Hang in there everybody. All right. Yeah, it’s awesome to be here. I’ve wanted to be a part of this presentation probably since its inception. Duann finally bullied me enough until I showed up. So yeah, it’s really humbling to be among so many talented individuals. A lot of great presentations already today. So my name is David Burpee.

0:27 If you didn’t already know that I’m currently a senior computational designer at Brooks running company I work on the innovation team so working with our biomechanics researchers engineers and developers to design running shoes of the future I also have my own computational design practice for the past 3 years or so which is primarily the work that I’ll be showing today I dabble a little bit in generative art and I am a lectur at the University of Washington for a couple of master of architecture programs sorry for the master of architecture program for two classes and I’ll talk a little bit about that towards the end and some socials there if you’d like to get in touch.

1:08 So I consider myself to be a multi-disiplinary computational designer. So just a little bit of kind of insight into my world. These are a few of the brands that I’ve worked with in the past. So everything from automotive to footwear some medical products, industrial design, and kind of flavors of a little bit of everything. And as Duann said, I’m here to talk to you today about time.

1:30 Maybe you thought this was a design presentation but specifically I want to talk about the past, present and future and how I believe that I think there’s kind of a convergence of some methods of manufacturing, material science and computational design that are beginning to enable some fairly interesting things as we look into the future. So speaking of past, my past is in the world of architecture. I was a lead high-rise designer leveraging computational design in the design of high-rise and super tall towers.

2:06 I lived and worked in LA for a few years before moving to Seattle where I did some projects there as well. So I have projects kind of all up the western seabboard as well as a few in Southeast Asia. And I was also always interested in kind of these generative and emergent systems. So I learned Processing kind of early on in my professional career out of school.

2:31 Which if you’re not familiar is kind of an IDE and programming language. I think it was developed at MIT. Shout out to the MIT folk here. And I just kind of loved building these dynamic systems that would control, you know, flocking algorithms and things like that. And it just felt like such a creative and kind of expressive way to use coding and design. And but I didn’t really know how to apply it to my own work.

2:58 So, I had these kind of two disperate interests when I began my career. I had this kind of like hardcore rigorous approach to thousand foot tall towers and designing, you know, 20,000 panels or something for them. And I had this kind of very creative more generative and exploratory side. For the past 3 or 4 years though, I’ve been primarily focused on footwear design and specifically using a tool called Houdini in that space.

3:23 I’m not the only person that utilizes it. I’m sure we have a few years users in here that use it for physical product, but it is not very common to use in that way. It’s primarily it’s a procedural tool that’s used mostly in film and VFX. But I’ll walk you through a few ways in which I use it in footwork today. And Houdini being a dynamic procedural environment allows me to use simulations, animations, kind of algorithmic design, in ways that I think are are fairly interesting and maybe offer an additional dimension over what I would consider like traditional computational design.

4:03 So for me it really allowed me to kind of combine my two interests, right? This kind of like computational design of physical product but also this kind of generative and emergent exploratory way of doing design. So I would be remiss if I didn’t talk a little bit about the AM of CDFAM given that we’re here. So I’ll show you a couple of projects kind of in this space.

4:27 You know like has been mentioned before like any good little computational designer I wanted to make some lattice structures. So I built out a a rig in Houdini. To create some lattice geometry. And the concept is fairly simple, right? I wanted to be able to take an input geometry with associated data, in this case, pressure map data, generate a graded like a performance graded lattice structure with variable density, and then within the actual lattice typology, have kind of additional layer of adjustability to vary beam width or wall thickness and things like that.

5:02 But I didn’t just want to make one type of lattice. I wanted one procedural rig which I could actually pick from a variety of different lattice typologies. So we have kind of the greatest hits of lises here that we all know and love. The conformal kind of beam and node lattice, the stochcastic voronoi lattice with variable density TPMS structures. And then kind of a space optimization algorithm there on the right.

5:25 And these are all performance graded in some way either by varying again the the density of the pattern itself or the beam or wall thickness or in some cases both. And these leverage VDB geometry which allows a for really fast compute and b once these go to mesh they are single continuous watertight services that are additive ready. So a little bit of an insight kind of under the hood for one of these setups.

5:57 This is the conformal lattice setup and I know the middle looks a little scary but the control mechanism is actually in the top right. It’s really just three kind of parameters that I am looking at and because it’s procedural it’s consuming like any input geometry that you give it and then really for the conformal lattice I’m just determining the kind of subdivision amount like the density of the lattice structure the number of vertical layers and then the unit cell type.

6:16 And so you see the kind of 10 or so unit cells there that I can pick from. And then the output is a nice conformal lattice structure that’s ready to print. Also being in Houdini, I thought, well, might as well squish this lattice structure. So just kind of a visual simulation of what that looks like under some deformation. And once I had this tool built, I wanted to put it to use.

6:46 And it allowed me to really kind of quickly iterate through some design options for either additive or hybrid manufacturing footwear. So this was a concept I did where I wanted to create a unified generative pattern that would be applied to the upper, midsole and outsole. So across all kind of three major components of the shoe. So again kind of straightforward process. Input the geometry data. In this case again incorporate the pressure map data, interpolate that into volutric data and then use that volumetric data to vary density and wall thickness variation and then output a warped gyroid TPMS structure.

7:23 And I’ll show you what I mean by a warped gyroid structure in a second. So this is a little bit of a peak into this setup. The kind of control mechanism is there on the right. So sliders and graphs, you know, things that we’re all familiar with. Some custom code that I wrote in a programming language called VEX which is native to Houdini and the geometry there on the left.

7:45 The key among you might see that there’s formulas for all the different you know kind of flavors of TPMS structures there. So Schwarzd Schwarz P diamond etc. So you can just kind of pick from those as well. And then I wanted to show this video mostly just to show kind of the speed of compute when you’re working at a volume or VDB level. So this is basically computing in real time.

8:15 And once the setup is built, I’m able to kind of continuously vary these TPMS structures, warp them, twist them, scale them, change their density, rotate them, just manipulate them in a in a variety of different ways, and then eventually just compute them downstream into mesh geometry. And again change the different kind of TPMS. This is I think the neovius which is definitely has the coolest name. Anyways, and then vary the density of this one as well.

8:47 So then just showing you know kind of the cross-sectional data through those volumes and how it creates the corresponding change in the wall thickness geometry. And you can also see the kind of single continuous surfaces in this view as well. Likewise for the outsole I took the TPMS pattern and projected the 2D geometry to the outsole and then again used pressure map to drive certain regions of the outsole geometry.

9:09 So where we have high pressure zones, it’s creating this kind of solid areas. Where we have lower pressure zones, it’s usually just kind of creating the outline of the pattern. And then that that white arrow you see is something that we sometimes use in footware design. We call it a gate line. It’s kind of this like theoretical line that we use to kind of flow geometry through and create motion and movement.

9:32 So creating this unified pattern between midsole and outsole. And then you can see kind of the the density variation in the midsole on the right. Especially from kind of forefoot to heel. And then this is just kind of a pretty image. It’s after you compute the volumetric values. It’s the values that then drive the wall thickness generation downstream. And then again applying that same kind of pattern principle to the upper in this case in these kind of solid areas that would be expressed on the shoe.

10:08 Okay, with that out of the way, we can actually get to the meat and potatoes of the the presentation. So what the hell is computational morphagenesis? Morphagenesis is the biological process that governs the development of an organism’s shape and structure. So it’s a process that informs the shape and structure of something. So keep that in mind. And speaking of time, I want to step back in time a little bit and talk about some of the early kind of generative approaches to digital computing, design computing.

10:44 So this is a piece of artwork by Dan Cooper. It’s called Triangulation 2. This was done in 1983 on a 4×7 in Apple monitor screen. It’s always a good reminder when you start to think that you’re doing something novel to remember that some dude did it on an Apple 2 like 40 years ago. But he had this quote that I really liked and I think is pretty profound today.

11:02 And it says the programs that he would write had no end point and could run indefinitely. He says he would watch the changing image until he saw a composition that he liked. And that’s really interesting, right? We always think of design as being this kind of like finite process where we’re searching for a resolution. We’re searching for an answer. And this idea of like letting something just kind of emerge and being able to use time as a design element is pretty interesting.

11:27 So, show of hands. Who here uses a 3D software for design? Okay. Now, keep your hand up if your 3D software of choice has what you see on the screen here, which is a timeline. All right, a few of you. I don’t mean like a history. I don’t mean like Killa, like a tree thing. I mean like a timeline you can scrub through. So, there are a few of you.

11:54 I would argue that this is one of the most interesting features of a design software because it literally gives you it’s cliche to say but an additional kind of dimension of design and it allows you to actually express time in design and time is something that we all deal with from day to day. So I’m sure we’ve all also made something that looks like this. Hopefully not as bad but some might say like well this is a really sophisticated system.

12:23 It’s got a lot of, you know, complexity. Obviously, there’s however many, you know, variables and parameters and sliders and it might output something fairly interesting. But at the end of the day, really what we actually experience is this. The thing that we make is inherently just a very narrow slice, a one set of those values that we actually express to make an object. So, does anyone know what this makes?

12:47 Shout it out. I’m not going to call on you. Good, good guess. So, no one except for him was going to be able to guess that it makes this but it would be difficult to know that if you just showed this unless you’re brainiac like this gentleman. So the the system inherently has no kind of context. It doesn’t actually inform the you know anyone but us what we’re actually doing.

13:17 And I always thought that was kind of a challenge, right? We make these systems that we spend so much time with finessing and perfecting and then eventually gets critiqued and goes into into something gets made but the beauty of the system is is lost. And so in a procedural environment if you are designing over time you actually have a series of these values that you can scan through and all of those parameters can be interrelated.

13:48 Some can control each other, some can consume each other, some might go away, some might be added, divided, whatever. But it provides you an additional kind of layer in which we can actually inform computational systems. So I’d like to talk about one project that I think exemplifies this usage of time. This is the EQLZ 360 performance basketball shoe and it was designed by Brett Gooliff with creative direction by Aaron Cooper.

14:12 And the brief for this project was creating a shoe that was really for the shifty and dynamic player and the person who kind of exemplifies 360° movements. And so we wanted to create a traction pattern that kind of exemplified this from a performance and an aesthetic standpoint. And we wanted to use some kind of natural algorithm to do it. The internal name for this project was the Duma, which if you know Swahili is another word for cheetah, which you see in the background here.

14:50 And so we always had this kind of natural spin that we wanted on this project. And we landed on the reaction diffusion kind of classification of patterns. So I’m sure many of you are familiar with. Thanks. And some of you have maybe even used before. But they’re fairly interesting for a couple of different reasons. Number one, they exist in nature in a variety of different plants and animals and things like that.

15:11 And number two, they can be described with actually a fairly simple equation that you see on the screen. One method is called the gray Scott model. And there’s really two values of this equation that allow you to create variance in patterns. The f and the k or the feed and the kill rates. So by feeding the the equation different f and k values, you can actually elicit very diverse range of pattern typologies.

15:36 So these are just four here. There’s a plethora of them. But you can see they sometimes grow at different speeds. Some kind of oscillate between two different forms. Some find a kind of equilibrium fairly quickly, but they’re these fairly dynamic systems that you can create with just these two kind of variables. And I’ve seen other reaction diffusion setups that have maybe two or three different patterns, but I didn’t actually know if they could be continuously graded.

16:05 So this was one of the early prototypes I did just on a rectangular piece of geometry. Where I just varied the the sorry the f and the k values across this geometry just to see like would it work? Would it create discontinuity? And I found that it it mostly worked but you can see that some areas don’t get filled in. It’s not quite perfect. So I needed a little bit more finessing.

16:31 And we wanted to basically take again these kind of high pressure zones and we wanted to map certain pattern typologies like this to certain zones on the outsole where we need high traction, high energy absorption and then allow other zones to be a little bit more disperately patterned so they can be more flexible and have lower weight. This would contribute overall to a more performant outsole. So I built this system that can consume this pressure map data.

16:52 It can use a process called algorithmic gradation which continuously varies takes that pressure map data and interpolates it to continuously vary the F and the K rates across the entire outsole geometry and it creates this really interesting kind of emergence system that creates a performancedriven pattern directly on the geometry itself. So you can see the outsole here just in a static view and some of those pattern typologies and how they correspond to different zones on the shoe.

17:23 And one of the challenges with the with a process like this, I sometimes show people that simulation, they I think they believe I just like hit the reaction diffusion button and send it off to tooling, but it actually requires to have a fully proceduralized system. And it requires quite a lot of of setup and control to kind of wrangle this dynamic system to actually at the endstream be able to go to tooling geometry.

17:51 So these are representative of a series of kind of variables or sets of data that I’m using to control certain areas of the pattern, control its growth rate, how it affects like boundary conditions, things like color dams and some of these punch through areas in the midsole. And then at the end of the stream on the right this is actually creating a 1:1 tooling geometry. So the simulation that you see here is not just visual this geometry is actually going and being used for tooling.

18:26 So again given this is a physical product and using a VFX tool we have to ensure that there’s some dimensional tolerances and things that we’re hitting. So we have half millimeter web thickness 4mm overall traction pattern. There’s certain features like the kind of zero fade up the sidewall and again the kind of responses to the different color dam and other zones on the midsole that had to be accounted for.

18:45 And so a couple of detail shots of that and you can see kind of how that pattern respects some of those boundaries, how it fades in certain areas. How it kind of dances around some of these spots and all the while kind of maintaining its ability to be manufactured. So hitting kind of dimensional tolerances and material dimensions as well. Once I had that procedural system built, it was really controllable and, really useful.

19:15 So, we wanted to basically create a series of launch animations. So, I built these animations prior to the launch of the shoe. So, the first one you saw used kind of the 360 logo to drive the pattern. This one is kind of an abstracted form of the shoe with the pattern kind of growing on outside of it. But it was really interesting because you kind of get this stuff for free.

19:35 Once you build this procedural system, you know, I don’t have to go back and say, “Oh, I want to like animate this now.” because it’s an ingrained part of the creation process. All this is stuff that we can just kind of showcase. So, you go back to that Grasshopper definition and it’s really difficult to decipher what it’s actually doing. But I think when you show people and you can actually sit with them and like watch it grow and watch it change and watch it morph, the system kind of becomes self-evident and everyone can kind of understand in the room even if they’re not a computational designer how it’s working and what it’s doing.

And lastly, the proof is in the pudding, right? So you know, this wouldn’t really matter unless it performed well. And there’s are a couple of wear testers that review the shoe very well. I obviously again can’t take performance or can’t take credit for the overall performance of the shoe, but I would like to think that the computational design maybe had a small contribution to at least the kind of outsole performance with these reviews.

20:32 So then quickly I’d like to talk a little bit about engineered living materials. So as I mentioned earlier I I co-e two classes at the University of Washington in the architecture program the master of architecture program I co-e them with professor Gel Lash and teaching assistant Kinsey Drake and this is an NSF funded grant project amongst three universities with a really diverse kind of range of of interests.

21:05 But our classes look specifically at utilizing living materials in the built environment and what it means to design them. How do you validate with them? How do you create these things for design? So for those of you that don’t know, engineered living materials are embedding engineered cells, living cells into 3D printable polymers which then elucidate certain dynamic behaviors like self- strengthening, self-hening, biorediation, filtering out toxins, biosensing and a host of other fairly interesting dynamic processes.

21:35 So this is one example that I think is pretty interesting. This was research done at the Nelson lab at the University of Washington. And it’s strain learning in proteinbased mechanical metamaterials. So these are ladder structures that are printed with living cells that as they are exposed to stress they actually selfh harden in the areas where they experience stress and strain. So we talk a lot about programmable materials in computational design, but this is an example of a material classification that can actually selfoptimize after it is manufactured in situ.

22:14 In its application which I think is pretty fascinating. And then a few images from the student work. They’re dealing a lot with you know various lattice typologies. And many of them were trying to explore this notion of like how these things will change over time. And I think we found that the typical kind of CAD approach methods and tools. It was really difficult to like start to communicate how these things are start to change to try to you know model their behaviors and things like that.

22:41 And we found that it was really not sufficient. So, I’ll leave you with this. I believe that some of the same kind of processes and approaches that were used by Dan Cooper in the early ’80s and some of the same kind of approaches that I used in Houdini to generate footwear are some of the ways in which we need to look into the future in order to design with these living materials that can change and create these complex systems over time.

23:10 And I would urge all of you to not underestimate the power of time in your designs. Thank you very much. To see the full recording of this and previous presentations, as well as information about future CDFAM events, visit CDFAM.com.