/sci/ - Science & Math

[Anonymous No.10010523]

Hemlo frens, i have next to zero mathematics backround and i am a brainlet. i've been messing around with simulating bipeds for fun, and i'm wondering what will help me understand rotations better.

for a joint to maintain a constant angle relative to itself and a surface (an ankle joint for example), whats the best way to avoid massive osscilation and feedback loops?

so far i've tried adding angle differences with a tolerance, and messing with Delta angles, but both are finnicky and i have no idea what i'm doing.

what is the most efficient, stable way for a simulated joint to maintain a desired target angle?

i feel like i've come close a few times but i'm more less fumbling around in the dark.

[Anonymous No.10010599]

>>10010523
I have no idea but Ill bump your thread, fren.

[Anonymous No.10010626]

>>10010599

Thanks, fren.

have a productive afternoon.

[Anonymous No.10011396]

please halp

[Anonymous No.10011908]

I'll give you a late night bump

[Anonymous No.10012828]

last bump before i go somewhere else i guess

[Anonymous No.10013043]

>>10010523
you're not going to be able to simulate a biped that balances itself by calculating and applying forces through its joints. This is hard and is the reason you didn't see walking robots until the last few years. If you just wanted to do one joint by itself, to avoid oscillation etc. look up PID control. But a pair of legs with 3 or four joints (knees etc) isn't going to be stable/balance itself if the joints are stabilised. You have to do the whole thing together which is hard.

[Anonymous No.10013073]

>>10013043

>you're not going to be able to simulate a biped that balances itself by calculating and applying forces through its joints.

i actually have, its not perfect but its rudimentary. i am doing the whole thing together. i have a variable stability recovery sequence thats uses rotation data, center of balance/gravity and other things.

theres a few ways i could do it without better joint stability but i'd like to go forward with this the "right" way.

[Anonymous No.10013107]

>>10010523
Rotation algebra and geometry are a solved problem. Go back to the 1800s with your outdated problems.

[Anonymous No.10013120]

>>10013107

i know they're solved, i'm asking which is the best
applicable solution.

[Anonymous No.10013126]

the math in biophysics maybe, or if you're ballsy enough, you can go for a home run and learn topology

[Anonymous No.10013185]

>>10013126

its actually not that complicated, like i said i can go ahead with my plan A if i have to but i guess i'll go read up on PID's

[Anonymous No.10014334]

>>10013185
give some more detail of what you're actually doing and you can get more specific advice (or at least let people tear apart why you're just 'cheating'). I mean are you doing 2D planar, using mujoco or box2d or what. Are you doing collision detection or are the feet stuck to the ground. Walking or standing still etc