Boid simulation (Collision teleporting boids?)

Hi. I got this particle simulation using boids (see the rule order on the video). They work fine.
I introduce an obstacle and all of the sudden boids start teleporting from their normal trajectory all the way up to where the collider object is. WTF?!
Please help. The boid animation should hit and fall from the obstacle.
Thanks.
https://youtu.be/Y32EH3VSF4w

Didn’t look at the video, but if you really want to have your file serviced (checked & fixed), provide it.

Thanks for picking up interest on this thread. Boid teleportation from Z- to +Z DIRECTLY on the collision object. Even if you move the object during playback at collision time, boids will teleport from flying to INSIDE the collision object. Teleportation at 0:57 on the video.
Are you a blend dev? I would like to invest time sending the file if I would know it would end up reported as a bug.

Are you sure they’re teleporting and aren’t newly emitted particles for some reason? When I’ve watched the video I don’t see any of the existing ones disappear when the new ones show up.

If you can post your file here then perhaps some people can look at it and confirm whether it looks like a bug. You can always submit a bug report yourself when you’re pretty sure there’s no other explanation.

I´m not reporting a bug. They don´t believe me. They always tell me to “proove” the error. Sometimes it is possible. OTHER NOT SO MUCH.

But I hope this file serves in some purpose. Thank you for your attention on this subject.
http://pasteall.org/blend/index.php?id=51101

Well, I can see the particles seem to hit a lower section of your collision object and then they show up (inside?) a higher-up part of it. I switched the particles to render as points with the Number option so you can track each individual particle to confirm they’re teleporting as you say.

Loading the file into 2.80 calculates the same and shows the same behavior (though scrubbing the timeline does not preserve the cache which makes it hard to investigate there, and trying to bake it caused a crash).

Maybe the physics system is getting confused by the complexity of the collision object. I wonder if splitting the connected sections into separate objects would make it happier.

Would be interesting to see if a simple particle emitter set up to shoot particles at the same face of the collision object would reproduce the teleportation without all the other animation, boids, etc. If that would do it you could submit that as a simple scene to reproduce the bug.

I´ve been in the industry for 17+ years. I know a bad software behavior when I see one. It´s broken. Namely by the devs, particles system need so much care. I’ve been recording natural phenomenon on particles since 3 years ago by the software simulation. I got footage that will spook your socks off in a middle of production. Yes, I´m talking 2.79b

But thanks again for your interest. You´ve been nice with this.
Like I sadi: I´m not reporting. Wish to have the time to deal with the blockiness of the general response of the devs in these issues DIRECTLY on blender bugs (we’re talking about 2.79b -again).

Cheers.

oh, btw: If you create a plane, rotate it -15º so that it faces against boids, the same behavior will happen.
It´s bad logic.
AI badly coded.
So the collision object, (even with friction and trespassing rules), will not yield to switch the boid to “fall” or deviate).

No bold one dares to adventure into boids with collisions and ai? Not collisions to land, but to FLY.
See how those nice instances will teleport themselves better than Tracer.

There are a few things you can adjust to make the collision better. The teleportation is happening because the collision object is not configured correctly. You are colliding with a bounding box (not the intended displayed object), then the next face (on top of the box) is used to determine the exit position. Prepare a higher density collision mesh and make sure it is above the collision modifier on the stack. You can add a collision rule to the fuzzy logic. Also boids need to always have a scale of 1.0 because that number is used to scale the collision sphere around the boids, too small and your collision become less accurate.
27_Teleporting Boids.blend (1.1 MB)
boid_woes

2 Likes

I did work subdividing this “collision mesh”. Results were the same.
The real issue is the teleporting boids. All thought the face /collision explanation seems to fit, it doesn´t explain why the next frame the boid particles should jump to a new “normalized” position.

I really thank you that you´ve taken the time to analyze this scene closely and that the issue might point to scale (in this we totally agree).

Though production time for this shot ended and I had a real emergency on this, shots were compromised and I had to delete incoherent boids by hand, masking them one by one on post production.

When you mention “prepare a higher density collision mesh and make sure it is above the collision modifier in the stack” <- you refer to the AI set of rules, right?

Thanks.