Agents up vector

[catchyid]

Hi,

 

I have a "spider" agent that I want to crawl against a sphere like shape, however Houdini always places the agent as if (0,1,0) is the up vector (i.e. like a human walking on a terrain). To fix the problem, I've 

-inserted "normal" node to my sphere to compute its point normals

-inserted "transform" node to my agent to change its initial orientation (copy sop node inside crowd source node aligns agents' z direction to template point normal N), so initially the agents look as if they standing on the sphere (see attached)

However, once I run the simulation, all agents switch back to a "strange" orientation (as if the transform I inserted has no effect)? I though that if I were to fix their initial orientation in the crowd source then it would propagate to the simulation. Also, I've added the sphere as terrain to the simulation hopefully to fix the problem but it did not.

So, to ask a specific question: How to make an agent align its up vector with the terrain normals while moving on the terrain? For human, up vector is always (0,1,0) , but for insects their up vector is always the surface normal...

Any feedback will be appreciated

Edited August 21, 2016 by catchyid

[cwhite]

There's an option on the particle motion tab of the crowd solver for how the agent's up vector can be optionally adjusted by the crowd solver. One option is to set it to the terrain normal.

[catchyid]

Thanks cwhite, this fixes the problem  

I got another a problem though. Basically, the up vector of an agent is only parallel to terrain normal, however it could be in the opposite direction, see attached plz. As a result, some agents are outside the sphere and some are inside. I've doubled checked that the sphere normals are always pointing outwards...

I've tried to set the up vector in crowdsource to match the sphere normal but that did not work (well, actually am not able to test it as I am having some difficulties copying sphere normals and pasting them to crowdsource up vector, I've posted a question on another thread...). Maybe if I can copy up vector to crowdsource correctly, the problem will get fixed...