Architectural modeling

david gary · Architectural modeling

So, after few chats with some users these days. Looks like a lot of people are interested in using Houdini as a procedural tool for architectural design.
Houdini is the perfect tool for complex building as well as for creating libraries of predefined and fully parametrable elements as assets.
Help to open the dialog between Houdini and the architectural world. What would you expect Houdini to do for you? What interesting features you use in AutoCAD or Rhino you think could be pushed to the next level with Houdini.

-d

aspect

hi david, me as an architect, there are 3 aspect we are very much interested in and wish to explore more in houdini:

1. generative component:
http://www.bentley.com/en-US/Markets/Building/GenerativeComponents.htm
as in bentley or rhino, we are very much interested in generating similar parts with specified variation as it grows along the grid or any other objects. This can also happen in designing city as well as building designs.
I explore such capacity in houdini with copy SOP> stamp> variable (which i link to other modeling SOP)> value (which I add different functions like $PT... $BBX... rand()...noise()... int()...abs()...etc.

Unfortunately, the houdini help doesn't cover much in this aspect especially on the functions part, if anyone knows a better method or more sophisticated method, would love to share more.

also wish i could find more in the DVD procedural city, still waiting to arrive.

aspect

2. L system

http://www.tropolism.com/2007/02/more_zaha_craziness.php#more
http://www.mathpuzzle.com/DaleWalton3.gif
http://www.theverymany.net/
http://www.theverymany.net/upload...Max_Test004_01_f_Pshop-705577.jpg
http://www.theverymany.net/upload...1228_SOM_Pavillon_05_b-710463.jpg

as an architect, we love to do fancy tiling, abstract or associated with natural pattern. However, the help in houdini has very limited resource on L-system. Hopefully we can find more at the DVD on houdini L-system.

aspect

3. Implicit form.

http://www.arcspace.com/architects/ito/taichung/taichung.html
http://rhino3.de/album/math/index.shtml

I think houdini iso-surface has such capacity to input implicit functions, however, again, the houdini help has very limited information... being an architect, our knowledge of mathematics function are very limited. some indepth tutorials covering such aspect would be very helpful

thanks!

david gary

Thanks A LOT !
These posts are extremely interesting: for our knowledge of your architrectural goals and for ideas of tutorials to develop in Houdini.

about generative components. That's all the copy sop ( not to say Houdini) is about! All of this is fully explained in the Cities DVD.

About Tiling:
An important pattern in tiling is Voronoi cells: as a matter of fact, this is also covered in the Cities DVD.
Tiling is a large subject... you can have different goals in tiling ( surface minimization, nice patterns like islamic ones... ).
About l-systems: l-system are for branching structures like trees, plants, lightning, cracks, bronchus, yet there exist « cellular l-system » which are totally different but allow some recursive tiling – like cellular division -( this will be covered in LsysDVD no4).

About implicit surfaces:
Actually there are not very good!! because they're.. implicit
1- this means they're of the form F(x,y,z)=0, and this is not the natural way for us ( for a human brain i mean) to represent surfaces as isosurfaces as the set of solutions of an equation.
2- They're not very precise, because the equation F(x,y,z)=0 has to be solved ( the set of solutions IS the surface), so Houdini solve the equation with numerical methods, so it's just an approximation + the resulting topology, especially the connection between vertices is not good.
( In math, all we care about is to turn implicit into explicit and to know where ( locally) we can re-write the equation F(x,y,z) into an explicit surface z=F(x,y) !... Implicit surfaces are rather a pb to solve and not a solution

as a matter of fact, that's not what is used in Rhino ( if i judge from the images you sent): what is used instead is parametric surfaces ( like Nurbs ( but with another parametrization, not only « rational »( NURBS = Non Uniform Rational Beta Splines) = (polynomial(x,y,z))/(polynomial(x,y,z), but with sin, cos, log).
This is achieved without any pbs with VEX or expressions ( and a little, tiny, knowledge of differential calculus, or let's say just math).

This raise the fundamental question, that you architects can answer: how do you want to have control over the surface: controlling a surface with CV or with evenly spaced isoparms(=curves) is very efficient and flexible and precise. That's why it is used so much.
Once again, i think the answer lies in either VEX/CHOPs or usual SOPs ( keeping the nurbs « type », but procedurally defining the isoparms for examples). Like « procedural stairs ».
What could be the other controls, or the constraints you would put on your surfaces, i have no idea: surface minimization, but this is WAY more complicated, and most of time, people used built-in functions ?
I wouldn't mind teaching differential calculus, but the most important thing is to find interesting applications!

aspect · thanks for the information

Hi david, thanks for the reply, very informative.

About tiling, I understand a number of software can do tilings in voronoi cells, however, some happens to be just shaders. For architectural purpose, it needs to be an object with a volume/solid lines for transferring to autocad for construction drawing purpose. http://www.theverymany.net/upload..._VT_test001_01_e_Pshop-742021.jpg
http://www.theverymany.net/upload...able_Test001_diagram01-706233.jpg
may be from a vex/shop to become a sop in Houdini?

Cellular l-system will be very useful for architects especially for urban planning, façade window design. Here are some references:
http://www.emergentarchitecture.com/projects.php?id=13
http://www.materialsystems.org/
I hope we will see the publication of LsysDVD no4, would it be very soon?

For implicit surfaces:
As I have very limited knowledge on math principles, I assume implicit function is written so as to formulate an explicit surface, when after several modification of the explicit surface had taken place, to see whether an explicit surface can be formulate into an equation. Anyway, you are right about achieve with vex/expressions, would be very useful to architect if there are some demonstrations of how to design such surface with Houdini, since these kind are applicable from furniture to building design, also as in tai chung the opera house to be constructed in Taiwan.
http://www.theverymany.net/upload.../070104_VARIABLES_L_03-729507.jpg
http://www.theverymany.net/upload...1228_SOM_Pavillon_05_b-710463.jpg
http://www.arcspace.com/architects/ito/taichung/taichung.html

for surface control: I think many use rhino for manipulation of prefer curvature, then transfer to autocad for precise dimensioning. I think the level of control in rhino would be sufficient for design, since we would do the final dimensioning in autocad for construction.
However, if we can set the degree of beveling/dimension of extrusion/the precise offsetting of isoparms/numerically setting the degree of curvature would be an ideal, and architects do many repetitions, is best to have more control in repetition with minor variations. Anyway, I think we are already used to CV control over the curvature we like.

Regarding the surface minimization would not be an issue, since there are many factors governing the construction process, we architects are more attracted to the looks than the economics.

I hope this information can help a bit, its best to consult with major corporate architectural firm since they are the one who produce and construct the most. One of the firm SOM http://www.som.com/opener.cfm they use rhino for design then autocad/revit for construction drawings.

Anyway, i would input more once i receive the DVD on l-system & procedural cities.

cheers

izno

This is a very nice topic.

Well I’m a newB in Houdini, but I’m trying to catch up as fastest I can!
As all the fine links by “aspect” is showing procedural design is one of the hottest topics of popular architecture today.

For me as Architect, becoming REAL is the most important part. How my dynamic design in my computer is becoming REAL and physical. Fore that I need to be more in charge of deciding the path of manufacturing to control the costs of the building. It’s easy to design without having to build it;o)

What would be nice is how to work and design in Houdini parallel with other software such as Rhino. Import and export is the most critical part as I can see! No doubt Houdini is the master as making things procedural and complex but in detail Rhino is needed for making it REAL when it comes to buildings and smaller scale, City planning and Urban design is not a problem.

This is an interesting project: ParaCloud http://www.paracloud.com/ showing how MS excel is boosting Rhino to a powerful parametric design tool. But the workflow and the workarounds are to complex to be universal tool for an office! The software is just in beta version right now.

I can imagine the combination of Houdini and Rhino would be an ultimate combination, if import and export would be explored.

david gary

Thanks a lot for your replies! Sorry i was on vacation.

About the path from idea to REAL. I totally believe that Rhino or AutoCad is closer to architectural needs, although i don't know how exactly .

Of course, Houdini being a multi-purpose tool, it has not been especially designed to fit architectural needs, and is more about conception. You cannot really blame him for that Smile

I think the best pipeline would be to keepAutocad (or Rhino) for the end of the process because it already has the proper tools for "REAL" export and that architects are more used to work with it. So Houdini would enter in the scene where Rhino leaves off in terms of complexity...
That really adds to my idea that we must release one or several other DVDs about procedural modeling, more NURBS-oriented and more buildings-oriented ( more than city-oriented).

But first i propose to start a little brainstorming area here to help Houdini meet architecture:

where i see Houdini can help is ALSO

-with its NON-DESTRUCTIVE workflow ( which goes with the procedural paradigm, but which
is NOT the same: it's not a matter of "multiplying" geometry anymore but a matter of working with the network ( like "the stack" but non-linear!) of your operations.
This means than you can change a parameter, for example the size of your window, and the whole result (like the whole building) or a geometry somewhere else updates.
I've played a lot with "procedural stairs" that automatically adapt to your terrain.
I imagine this is very interesting for architecture because you always have to deal with little modifications and constraints where you don't want to redesign everything and elements that work in relation with each other. For example:
a procedural stair which would be automatically modified if you change the size of your room.
The architectural elements "send info one to each other", so they need to
be thought and designed in interdependance.
Working with some modules ( rooms, window, stairs, roofs whatever)
who "speak" one with each other is very useful for the conceptual side of architecture
( make your space functional, dealing with ecological constraints, orientation, access ). I've had long discussions with a friend architect on that!

But it's also useful for the practical side: i want to change
one parameter but i want to have the same relation between elements.
I mean it can be turned into an "architectural paradigm"

-One great thing in Houdini, which can make me think it's also appropriate tool for the REAL part is that Houdini has a powerful managements of groups ( groups of geometric primitives):
for example ( lets suppose a tile of M m2 costs N dollars), you can create a group with all the tiles and have Houdini interactively and procedurally computing the total expense ( better than Excel sheets!), then you can send this in CHOPs ( the " abstract MATH engine") and make Houdini procedurally create additional elements if the price < certain value.
( you send your constraints back in the network). This is of course the minimum you can do.

Or you have builded several types of windows, and the proprietary says: too expensive! then you just
switch to another type of window in 1 click
( completely parametrable) and you immediately get the result!

Developing Houdini for architecture sounds like an incredibly interesting project, but we need to know more on which scale architects work. Some work on part of cities, some work on buildings, some work on ergonomy and design, on the "artistic side".. some work on standard flats etc!

-d

izno

Hi David,

I’ll try to put some more info into this forum to fill the gaps, sorry for the spread of info.

First of all I have to congratulate you; I have just received your DVD on procedural cities. Fantastic DVD, I wish I hade it for 4-years ago when I did my Master at the architect school. At that time I made a software prototype my self to work with a dynamic model of a city. With the knowledge I have today I could do much more... But it’s not to late;o)

Well I have to say your understanding of REAL is very precise and I’m impressed. The kind of tools you are talking about is really nice and in a near future every architect (at least her in DK) has to use it. New DK-government directions from 1.st of Jan 2007 are that all public buildings must be developed as a digital-project. This means that there is a BIB standard made fore how to communicate architectural projects digitaly.

Here is some Digital project tools/software,

Graphisoft Archicad
http://www.graphisoft.com/products/archicad/ac10/
Not procedural but it works with components and it gives you a spread sheet where you can calculate your data in form of door, window and so on…

Autodesk REVIT new revolution in CAD!
http://usa.autodesk.com/adsk/serv...ndex?siteID=123112&id=3781831

Autodesk REVIT, is one of the tools which will probably be a new standard here in DK. What REVIT dose is exactly what you are describing with calculating the cost, objects and much more at the same time you are designing a project, with the advantage of being CAD-DWG standard. The workflow is kind of procedural and component based. You are binding your elements to your geometry and by scaling, stretching and arraying your walls you get the building to grow, so on and so forth… (there is more to it…)

I think when it comes to it; HOUDINI is a fantastic tool to use for sketching developing and expanding your ideas in a more freeform and experimental way. Maybe programmatic, diagrammatic if you would like to call it, relations and the conceptual ideas could be explored.
I love the way HOUDINI works and the procedural modeling is fantastic. Of course there is always advantages and disadvantages in software packages, when it comes to it at the end it’s the man behind the computer is important and the tools are just tools… what you make with it is what it counts;o)

I’m really interested in exploring HOUDINI to its limits (even it’s hard to imagine HOUDIN could have any limits;o)

Just let me know if I can be to any help to describe what HOUDINI for architects could evolve into.

I have a wish list which is not so different from "aspect" wishes, I’ll try to gather some more info about it and post them here.

For now I leave it with great thanks for the DVD tutorial.
//

david gary

Very very interesting, thanks.

First, really it's never too late.. Plus, if you're « tech-minded », you'll certainly find Houdini easier to learn. I guarantee it!
because in Houdini, you KNOW what you're doing, and it does exactly what you think it does. That's the magic of operators. This means you don't have much pre-built effects like in Maya, for example where you click in a menu full of options, but you don't know how it's internally. The internal representation of the 3d workflow is the same as the external representation, there's no gap between!

REVIT seems to be very close to Houdini indeed, and interestingly to the ideas i presented ( the « tile example », their « bidirectional associativity », the library of components etc). I don't know if Houdini would be redundant with it
I continue on my ideas (that need to evolve of course), but this is how i see the things.
Houdini is very powerful but in your project you start from scratch and you can go in soo many different directions that you have to pre-organize your stuff and workflow, to work.Aa little bit like programming .So it has pros and cons, ( but at a high level enough where you don't have to reinvent the wheel everytime).
You have total control but on the other hand you have to specify everything the first time.

What i think is incredibly useful is assets: ( the DVD gives an example so i let the DVD explain): where you can encapsulate all your functionality in 1 customizable operator ( which represent ALL the building you're working on f.ex), then you easily create your interface with only the interesting parameters, update your textures etc. Like a big macro for your project that everybody can interactively play with ( without having to take care of what's underneath – Excatly like in object oriented programming)!
It depends on the flexibility you want to get and the time you spend on it.
What is missing in Houdini for the moment is things like excel sheets management for example but H9 new python integration may change this.
I will keep on posting ideas.

About Houdini limits: EVERYDAY i think i've just found the limits and then Houdini surprises me again.
-d

aspect · afterthoughts

Hi david,

I agreed on designing a building with networks of operations. Because every meetings with clients there involved design changes. However, every country and cities has their “protype” and each protype has there own criteria. Take Hong Kong for example, the residential towers has 8 protypes, flat size/window size/room width, depth are critical. The initial settings of relations of parts or criteria is very important and almost every offices varies. Most of the time, those networking may be too complicated for architects to comprehend, I picture that they may hire a Houdini consultants or a freelancer to set up their own library in the future.

The cost and sizing in Houdini will be a very preliminary reference materials for architects. Since cost/budgeting are done by quantity surveyor, they would be very grateful if u count the tiles for them, however, there are many factor affect the cost, most importantly is labour, being the ease of workflow is more critical most of time than quatity of materials.

I think it would be great to cover more on nurbs, since there are sops which works on nurbs surfaces only like creep/project/trim and these are very useful sops for archects.

I finished your tutorials and its been a great learning experiences and I already try some of the things I learn for my upcoming design, will forward to you when is done. The following are my afterthoughts:

The voronoise is very useful for architects and we mainly use it for organization, tilings and façade design. However, I had to admit that the vex for creating the center of attraction is quite hard for me to comprehend and most of the time we architects likes to work with multiple center of attractions like in here http://farm1.static.flickr.com/230/497557812_ace0ec5044_o.jpg . I also tried to experiment other pattern like cellular or Worley noises, may be I’m lack of knowledge for vex, many attempts http://farm1.static.flickr.com/214/497557808_265f6c80b1_o.jpg were not successful in making it into trace sop 3d objects. However, I can picture the vex would be a very powerful tool for architects in pattern designing and make it into 3d objects and to project/creep onto the different shapes of buildings.

The L system for geometry creation such as optimum packing with the magic number 135.7 are very intriguing. I can see this being apply to massing studies, the light distribution on urban scale.

There is a great architectural book on this issue, I think u can flip over it at the architectural book store. It shows the fundamental of algorithmic approach to architecture. And some of the topics also covered at your dvd also.
http://www.amazon.com/o/ASIN/1568...bscriptionId=108P28193HCTR2B9FAR2

here are more references site for algorithmic application to architecture:
http://www.futurefeeder.com/
http://www.processing.org/
http://www.proxyarch.com/
http://www.reconstructivism.net/

cheers!

david gary

Wow, that is some incredible feedback!

I invite everybody in the forum to explore your links. I've began already... awesome. I knew processing.org but processing remains a scripted language and i would like the world of 2d graphics and vector to share the same viusal node-based approach as Houdini, like Node-based flash design. Actually you can almost do that in Houdini, but there's no real vector export!

About Voronoise, Voronoi cells. Indeed, to have artistic control, you MUST define the cell centers first.
Actually this is done by the VEX operator, but this is hidden in the code. It spreads points randomly but uniformly on a surface, but this is a process you can override in shading for example with point clouds. Btw. If you have any questions about specific points in the DVD that remain unclear, feel free to ask!

About the magic number 137.5 in l-systems, yeah funny. I speak about it in the Lsys DVD, but i don't go into the mathematical explanations, but there's really interesting math behind. Nature is full of optimization, minimization process, like minimization of space ( for packing, or maybe the most spectacular is soap bubbles surfaces that materially resolve the problem of minimization of surface between two curves!). I can imagine how this can be interesting in architecture.

The book seems very promising i will give it a glance if i find it.

About Nurbs: yeah, i feel that it was a little left aside and that it deserves a full DVD, because you can achieve wonderful things with them and procedurally.

I can't wait to see what you've done with my tutorials. . Just post this in the Houdini WIP section.

-d

izno

aspect

hi david,

i just have upload some of my work at the W.I.P section, everyone please feel free to advise for further improvement!! thanks in advance.

My further studies which I would like to develop further from the dvd is-

-To explore other vex+trace sop like cellular, which is very useful for façade design. I assume, the best way is too use creep sop and paste the voronoise geometry onto a tube or box geometry. Some examples I found –
http://farm2.static.flickr.com/1054/576031932_1ad5c31513_b.jpg
http://farm2.static.flickr.com/1320/576031912_e7aabb6195_b.jpg
http://www.materialsystems.org/
However, attempted had been make but not quite successful with the trace sop of developing geometry from the vex. http://farm1.static.flickr.com/214/497557808_265f6c80b1_o.jpg

-More control on the voronoise like developing multiple center of attractions or even 3d voronoise. Shown at this webs and also presented in TOOLING pamphlet architecture.
http://www.theverymany.net/labels/Voronoi%203D.html
http://farm1.static.flickr.com/230/497557812_ace0ec5044_o.jpg
http://farm2.static.flickr.com/1019/576031862_d11a927e05_o.jpg

-The L system in application to architectural geometry, I know this may become a bit too complicated for newbies to explore. However, results would be quite promising as shown at this web –
http://www.mh-portfolio.com/indexl.html

- The magic number137.5 in L system is quite interesting to explore which also shown here & in TOOLING pamphlet architecture.
http://www.materialsystems.org/blog/?p=136
http://farm2.static.flickr.com/1297/576031876_d03fb8cbe4_b.jpg

Really looking forward the next dvd and hope to be happening soon... and if anyone have any tips or link that help me to understand better with my explorations, that will be great!

david gary

Well, thanks A LOT again. Your contributions are always very exciting. Smile

1)About Voronoi: a generalization of Houdini's Voronoi can be provided with Point Clouds. The difference is that you can define your own centers around which every cell is going to be built ( the standard way that Houdini uses is to create a 2d/3d grid, to place a point at the center and add some jitter to his position: this significantly speeds up the process of finding the closest points cos you already know that they will belong to the neighbouring faces of the grid, even if you don't know where exactly because of the jitter , so you don't have to loop over all the other points to find the closest ).
point clouds use this idea in that they « naturally », « natively » loop only over points that are within a certain radius from a certain given point, but with the advantage that the distribution of points is really arbitrary.

2)The l-system link is simply amazing. Well Lsys DVD no4 will be about L-sys recursive/iterative modeling, to expand the concepts of l-systems to Houdini l-systems to have totally scriptable l-systems ( which is not the case yet) and to have general recursion-expansion in modeling, so it will use the same concepts as those presented in the link, to animate the construction of a city.

3)I like these excerpts of the TOOLING book
I like their simplicty, their essentiality: « recipe for cracking, for packing ». I'd like to do the same for Houdini, like writing a « procedural pamphlet », where you have the recipe of « solving the problem by dividing it: the copy SOP », « doing the same operation with varying parameters for each object: VEX or copystamping », « sending info, everything is a channel: CHOPs », « geometry remain geometry, the SOP context », with a minimalist graphic style.
Apropos, though quite off-topic, do you know John Maeda, not an architect, a graphist, but graphists use to know architects and vice-versa . It's my favourite one ( i have to say i don't know a lot of famous graphists), i've written some sort of an « article » on him in french, about his ideas: this kind of minimalist and essentialist approaches reminds me of his works ( some of his were oriented toward the concepts of « reactivity », « reactive media » ).

4)Packing... i wanna say a little word about packing, like packing spheres. The « recipe » shows as simple it is that complexity arises with iteration, well here again, Point Clouds are useful.
You spread points on a surface, then you create circles around each point, their radius must be such that they don't overlap. So you find the nearest neighbour with point clouds: divide the distance by two, this gives you the radius of the circle, do this for each point. Your surface is partially covered, then spread points on void areas and continue. If you change the type of distance, you can do the same with squares. i will show this in the upcoming Vex DVD.

David

aspect

hi david,

congradulation on your new dvd, i just have received and very excited. hopefully i'll try some new magic and report to u later.
thanks for introducing John Maeda, i do like his work for its simple graphic but highly sophisticated implications, i found it very architectural, would try to achieve such effect in my work also.

here are some interesting architectural link i discovered recently for your enjoyment-

http://www.dataisnature.com/
http://www.processing.org/
http://ncertainties.wordpress.com/blog/
http://www.flight404.com/blog/
http://en.wiki.mcneel.com/default...cNeel/PointsetReconstruction.html

cheers herman

izno

if you like Maeda:

http://blog.ted.com/2007/09/john_maeda.php

david gary

Hey guys:

Aspect ( Herman):

Thanks for all your links. Very interesting as always.
Shapes in nature, morphogenesis are a renewed wonderment each time i saw them! When you see plants, cristals, seashells, patterns in nature, you really see that the nature is doing math, or inherently following mathemtical rules.
I say a little word about that in VEXSOPs2 dvd, about seashells: not only the shape of seashell is extraordinary but also the mechanism that produced their textures (even if the textures themselves are not always spectacular): these are "purely morphological, or morphogenetical" chemical reactions, which means that their equations are completely independant of the nature of the chemical substances: geometric chemistry if you want...
and this is "usually" implemented in CG with cellular automata: a set of cells with extremely simple behaviour that you let evolve in time.

I pretty much liked the info on point set reconstruction which is a universal task in CG: voronoi, delaunay, voxels.. Well put here.
I'm glad to have introduced you to Maeda: i think you have watch the link Izno posted. Quite funny guy!

izno:

thanks very much for this video with Maeda: i have read him, i have listened to him, but i never watched him on what we can call here a little one-man-show. I didn't know he had this kind of humor. It was quite funny ( and with the drums on the side: you really felt like sometimes someone is going to hit a "bu-dum-bam" when he says something funny).
Above all, he really likes to play: i mean if you take the example of "painting with food" ( just visually): well that's not something new ( PaintFX in Maya is based on this idea), but what really interests him is to play with the thing.
The show itself is more about him and his irony about life, about the MIT and japanese food! lol

-d

aspect

Hi David,

I enjoy very much with the VexSops2 dvd especially discovering the “compile vex code to sop type” which are very useful for modeling. However, I’m a newbie in vex and some of the procedure and vex code are quite deep to comprehend and need some time to digest. U mentioned about the cellular automata in the DVD which is what I’m eager to explore and visualize in sop, the same with the other function like the Worley noise, voronoi noise and cellular noise. There were many attempts to try it on my own but not quite successful. Wonder if they will cover in the upcoming DVD? Or any quick tips in this regards?

Thanks.
Herman

Architectural modeling

thanks for the information

afterthoughts