I am a beginner in Houdini, I know chop is a very powerful section in Houdini. Dose anyone ever use chop to make some AI scene successfully. Also, does anyone know if there is any AI houdini tutorial? “Massive” is a bit too expensive to buy. That would be great if Houdini can archieve the crowd and AI effects.

thank you for your attention

Houdini would only be the “wrapper” for something like that. Yes, you could use Houdini as a tool to create a crowd simulation(if that's what your after), but most of the work in something like that is the implementation. Houdini won't handle things for you automatically - you'll need to do all the clever bits, incorporating the theory and all the incredibly tedious mucking about in code and OTLs(which would seem to be handy containers for a crowd sim). I'm not trying to discourage you, but you seem to be asking if there's building blocks for crowd sims already in Houdini, and I would say no. It would be a big project. However, assuming you don't want to write a complete environment to do it in, I would say Houdini is a good, flexible off-the-shelf place to do it. Lots of accessibility.

Cheers,

J.C.

I wonder how smart Caleb's mice were in Cats & Dogs and how much custom code he used.

Mailing list post:
http://www.sidefx.com/community/interact/mailarchive/index.php?view=WEB&msgid=405120AB.5010601@rhythm.com&perpage=20&revdate=off [sidefx.com]

Oh yeah. Crazy Caleb. :-D He has a lot more fascinating stories.

By the way, it'd be really awesome if Fuzzy Logics is implemented into Houdini in someway. FOP? May be??? Oops… I think I am drooling here…



:wink:

Well, technically, you should be able to do anything via events in the Logic CHOP …

(for the tech minded, you should be able to implement a Turing machine)

Hey,

I have already done some basic crowd sims in Houdini so yeah you can do it. I used particles in POPs mainly in teamed up with expressions in SOPS and CHOPS. Then in VOPS you can assign a different shader for every instance or animate the shader values to be different for every instance for every particle character you generate. You can in addition add coding expressions to the POPs to add behavior chains for the AI. Houdini is well setup for it. If you are going to siggraph this year they have a crowd sim course on wednesday that I suggest you attend. I have found that you want to use instances for Crowds in Houdini. Copy geometry is nice but I got about a 3 month render time with the copies and about a 1 hour render with the instances. I am not kidding you. There is that much of a difference. Anyhow the buzz videos off of his site you can either go through the course or just buy them. I did both. I suggest buying them. After go through those learning videos and the ones from sidefx then you want to check out and read all of the pdf tutorials on Sidefx's site. This one in particular http://www.sidefx.com/community/learn/effects/legacy/POPs/pops_flocking.pdf [sidefx.com]

Cheers,
Nate Nesler

P.S. Hopefully I will post some of my results soon as I am getting ready for siggraph.

Thanks! matrixman!

I will go through that Houdini flock tutorial soon.
But I think using instance object in a crowd scene is not very effective. Because it is very difficult to blend motion curves (especially when behaviour changes). I think I would use renderman 's Level of Detail technique to solve render problem.

Hey,

Well I read somewhere that massive uses instances too, and now I can't seem to find where I read that. So don't qoute me on that until I can remember where it was I read that. I just don't quite understand how you can have individual animation for each instance. I though deformations were transfered through all the instances because every instance was based off of the memory allocated definition of that object. So if you change that definition for one, every reference will take that new and different object. I am waiting to find out how they are doing it when I am in the siggraph course. Unless anyone happens to know which I would find very interesting to say the least. As far as blending motion curves you have to blend between particular stances. For instance a fighter will do a move and then come back to that stance before moving to the next move but when put together the motion becomes very fluid. If you have Lord of the Rings Volume II special 4 DVD set then they even mention this and illustrate it when speaking about the motion capture and how to put the moves together. Thats how I would approach it.

Cheers,
Nate Nesler

MatrixMan,

how was the crowd simulation course in Siggraph 2004?
I found “Massive” is definatly a trend for future AI development! This is because if I use Houdini to solve the AI problems, huge amount of time is needed! (Especially doing crowd with cloth simulation at the same time.)

I am looking forward to see your crowd sequence which is simulated by Houdini!

Hey,

Well I found my system was remarkably like ILM's and PDI. Although I don't have the banking feature. I have kind of come to the conclusion that it would be best to have a mix between the two systems. Massive for up close shots and the ILM, PDI, My System which I have not named anything yet for distance shots. lol The reason I say this is that there were times in massive when they had to write and figure out all these complex rule sets and what not just to do an extreme long distant shot that I know I could have done with mine in like 5 minutes. Thats not render time mind you I just mean setup time for it and I would have gotten the same results. Also you can not really control massive you can only make suggestions to the characters they do what they want. In times when you want complete control ILM, PDI, and My System make more since. I am going to implement a LOD in mine and start baking because I had already come to this conclusion but the course verified this need. Yeah Bake Bake and Bake some more. I will have to say the implementation of doing a crowd the ILM Particle style is easier in Houdini and is more user friendly for working with it than their implementation in Maya where they had to write lots of code to do simple things that Houdini does right out of the box not to mention that they had to bake the simulation in order to delete characters post simulation which left the problems of character avoiding and interacting with a ghost space. In houdini you could just delete this character and rerun the simulation no problem. I think a Hybrid system between particle based Crowds and maybe Massive. I want to keep mine with the houdini package and write independant plugins that will work with houdini instead of creating the wheel from scratch. Seems like alot of extra work that may not be required to get the job done.

Cheers,
Nate Nesler

Is that Flocking pdf still available somewhere?