Slime Mold Simulation. Experiments.
ฝัง
- เผยแพร่เมื่อ 18 ก.ย. 2024
- The algorithm is known by the names "Physarum" or "Slime Mold Simulation". Numerous agents rush across the field, leaving pheromone trails. They prefer going where there is more pheromone.
Since the algorithm is based on the agents interacting and generating beautiful patterns, this is a way for me to experiment in areas such as swarm intelligence and generative art.
You can support the channel on Patreon:
/ simulifehub
More support, more opportunities for new projects.
I would love to see the tricks where you made your logo appear from the slime mold become your new intro animations.
It is entirely unique to you, and you deserve it!
I assumed it was just reversed?
This reminds me of ants too. I think why it looks more like ants than slime mold to me is that slime mold would also leave tendrils of where they "died"/failed to find their food source and tends to make delta-like tendrils instead of tiny agents leaving pheromone.
Dont a lot of slime molds just kinda like 'slurp' their tendrils that didn't find anything, back in? Im not sure
@@astrovation3281 Yeah, they do. They retract parts of themselves back into the mass to prevent biomass loss.
@@ZerglordGhi So i guess the simulation is more accurate this way, right?
Totally agree with you.
But the algorithm is so called Physarum
I am not guilty :(
What about some agents randomly becoming "evangelists" or something? Their parameters are slightly tweaked, and as they move around their pheromones tell other agents to become more like the evangelist. Eventually something will get selected for and the whole system would change. It'd be interesting to see if the system converges to something quickly or if it constantly changes.
I'll think about it ))
I love the sounds you applied to the simulations, it really shows you know how to present a program as art 🧡
If you made this with a genetic algorithm, the metric i can think of is "followed pheromone trail the longest", either by rhe longest stretch it sought the pheromone, or the total time it spent seeking the pheromone. It would be interesting to see what kind of pattern it would settle on once the population becomes homogeneous.
I would rather have had more information on which changes illicited which results. The Settings are what make it interesting!
The angle of the sensors almost acts as a surface tension, the wider the angle the more easily they can be influenced by other agents, and more specifically "in" further into high/low regions depending on mode.
Wobble is analogous to temperature rather directly, while the angle of movement is analogous to mass, as an agent can only be diverted from its current velocity vector (forward) by a certain amount per unit "force"
Hyped for more Simulife Hub content!!
I think the global parameter changes should start with a random ant and spread like a disease. So there are always 3 sets of global parameters, 1 for healthy ants, another for infected ants, and an interpolated blend between those parameters, to animate the transition. It should spread through pheromones.
I also think the pheromones could be distorted in a separate process, like with displacement maps that ripple, flow maps, or masks that morph in,
Could you upload this somewhere so I could run it as well? This looks so good!
I post the code on Patreone for paid subscribers.
5:57 Maybe some light would be shed on why the clear areas form if you programmed in the ability to see which way the individual mold cells are all facing. I'm almost certain that around the edge of the bubbles all the cells are facing the same way.
In the program itself, there is no way to see the direction of the agent. A bunch of dots rush across the field.
To explain in the video, I separately wrote a simplified version with large agents.
But in the main program it is worth adding such a mode
Wish we could download this and run it our self, looks like i could spend hours at this
I didn't knew you've English version of your channel but I was surprised there's not your voice.
No experience in speaking English
Have a "disperse" pheromone
Reminds me of a mushroom trip
definitely should add a genetic algorithm!!!
Spooky intro!
What is the medium in which it grows?
?
love it !!!
I would have assumed this was running on the GPU (since it's obviously easy to see how it could be massively parallel), but I'm impressed you're running it this well on just the CPU.
At least, I think you said it's all on the CPU, right?
He could be using the AVX CPU instructions. Not as fast as GPU, but you can calculate i believe 8 floats X Core/Threads(not sure which) per cycle(okay not true but close enough).
So something like a Ryzen 2700 can do 16 per cycle, with a clock speed of over 3 GHZ
Usually such projects are made on the GPU and use more than a million agents, but I have no such experience
I iterate through all agents in a for loop.
With a field size of 1280X720 and 40,000 agents, about 30 frames per second are obtained.
But the GPU brings its limitations, I have more room to experiment
😘 Your voice, NOW!
What about including a behaviour that can be triggered where the agent reduce the pheromone in it's trail. Plus could you potentially in the future offer a version for people to download and play with.
The code is posted on Patreon for paid subscribers. I thought about your idea too. When I get back to the new version, I will add this
Love it
If anyone is experimenting with this & has access to a classroom, hav3 each student design an organism with several attributes & rules of behavior…
Attributes: fat reserves, quickness, likelihood of mutation, ability to recognize food & danger, if it senses that it’s being eaten, release a spore containing information about The organism that’s attacking it… + ?
Is there a reason why you don't share links to origial material?
I know this specific video is in Russian, but still it feels like a good sport to mention the original author
th-cam.com/video/hw_t77lseQs/w-d-xo.html
Thank you. This is my project and I control both channels, for the Russian-speaking audience and for the English-speaking
@@foo52ru
I see, I have no questions then!
It looked like someone was copying your work without proper credit, but if that's you - obviously there's no issue :D
Keep up the good work!
I wonder if you can achieve self-organized criticality with this?
That's what I'm interested in
@@wallcraft-video a guess a first step would be to look at what normal criticality looks like. This would involve finding sharp transitions in behavior relative to the parameters ("phase transitions") and then setting the parameters to be exactly on the boundary. Also see en.wikipedia.org/wiki/Edge_of_chaos
I am first :/
But not last
@@heavencanceller1863 Yes
@@r0fael_programmer Everyone is last for a short period of time.