#Struggling with ML-Agents
1 messages · Page 1 of 1 (latest)
Start from playing with the ragdoll itself. What parameters of what components affect the character ability to stand up? And how would you know if it stand upright if you couldn't look at the character visually, but only at the various parameters? Once you have that figured out, you would know what parameters to let the model control and what parameters to monitor to provide it rewards during training.
hmm this is solid advice but I sadly don't have the time for it 😦
Well, this is not a project that you can do quickly and without thorough research. If you don't have the time for it, maybe choose a simpler goal.
standing upright isn't a simple goal damn
It's not at all. Why do you think there are millions of dollars being put into developing humanoid robots currently?
seen a lot of projects on yt and github about it tho
Sure, it's a lot simpler in a virtual environment, but it's still a pretty complex task.
that's fairs
Especially if you don't have any experience in the field
Any documentation or anything at all I can read to at least nudge me in the right direction
true
rn it feels like i'm beating round a huge bush
like HUGE
As I said, research how a ragdoll works first what makes it move in certain ways and makes it stand upright. I don't think there's much documentation on ragdoll in unity docs, but it's basically just rbs with joints. So research how these joints work.
Then separately you need to research how ml agents work in unity.
Finally combine the knowledge to implement the thing.
Maybe use Google or YouTube to learn more about Ragdolls. The general concept is the same regardless of the engine used.
I know how the unity mlagents works it's just the ragdoll I never understood
It's all about physics. Think of a door. It can rotate around an axis when a force is applied to it. That's basically one joint in your ragdoll. The combination of many of these joints and forces/torque applied to them could make it move in certain way or stand upright.