has anyone else here worked on physically simulating bipeds? i'm trying to create a mostly comprehensive solution for my game and i was wondering if anyone here has also attempted this and had good results. i have a decent grasp of what i need to do to from here but any tips would be helpful.
i've read a few papers regarding this but i would be interested in hearing what /sci has to say.
Wish I could help, but this isn't in my wheel house. Have a bump, because I really liked the interaction between your bot and the box. The stumble/ off balance portions are done well.
Mason Myers
you need some serious biomechanical experties to pull off a convincing dynamic ragdoll. The only game that came close were gta 4,5. And they did that with help of on site engineer directly from euphoria middleware company.
Anthony Ortiz
i found some helpful proportional charts from boston dynamics to properly weight everything
Jacob Davis
last post, seems like not something a lot of people here are into.
David Rivera
nope but i think i could help maybe. if it's for a game, well that doesn't totally narrow it down but depending on what you're doing you could include a sort of physics breaker, maybe the feet are allowed to behave like wheels and can just move along the ground how they'd like or just a virtual string on their head to pull up on like a marionette
People naturally use the weight of their head and their upper bodies to guide their movements intentionally, while this is just the opposite. Pretty much an exact simulation of a drunk whose muscles no longer coordinate. I think the spine probably needs more focus. The spine should act just like a snake kinda. Curving movement starts at the head and travels down. Once curve that stops midway will create forward motion, a curve that goes all the way would lead to tumbling, a curve with a second opposite curve stopping halfway would lead to forward momentum that is then supported by leg movement, and a curve plus an opposite curve going all the way down would result in a slithering effect of leaving the upper body in a forward position but without forward motion. The curve could also start at the pelvis and direct upward
Alexander Butler
well that video in particular i loosened everything up to give him the impression of brain damage.
basically i'm planning on completely reworking the system to have each limb have a seperate self balancing behaviour based on target rotation data. i have to more carefully study how humans use the inertia of their limbs to stabilize.
Jaxon Martinez
but thanks ill try to visualize your suggestions with my current joint setup. i get that people kind of "flick" or "whip" themselves in waves of motion depending on the force and direction. my goal is a totally dynamic solution that leads that responds as organically as possible with as few cheats as i can manage.
Joseph Russell
This is great. I have been very interested in kinematics for a long time, but haven't delved into that rabbit hole yet. I would look at how bipedal hobby robots are programmed for some ideas. You already mentioned Boston Dynamics, but check out how some of the Robo-one bots walk, as well as things like Hovis, Bioloid or Darwin walk. the trossenrobotics forums are a great place for this. Similar to your project; a game called "Sub Rosa" simulates walking (or at least seems to) much like you are doing. Check the developer's blog. In any case, keep posting webms. This is a very cool project. Keep up the good work!
Lincoln Gutierrez
Just simulate a car, set half of its visibility to false, and replace the wheel images with feet images.
ill check out the trossin robotics forums. i'm aware of sub rosa, i became aware of it after i started my project, but my game is going to have more of emphasis on melee combat. intially i started with some very basic simulation of things like the klann linkage and theo jenson but i knew that for what i had in mind i needed organic looking somewhat human proportioned figures. im trying to find that balance between super unnecessary simulation and fun movement mechanics.
Jaxon Nelson
this is actually a good reference video lol. i'm trying to find some decent footage of people falling over or stumbling while running to break down all the minutia of that process.
Nicholas Wright
yea but thats boring. i want WWE with clunky garbage robots made of garbage.
>is your box person retarded. Are you incapable of reading a 15 post thread? >well that video in particular i loosened everything up to give him the impression of brain damage.
Charles Nelson
i didn't read that post yet calm down lol.
Angel Cooper
You could've taken the ten seconds required to read the thread before asking a question.
Jaxson Bailey
Check out some of DeepMind's papers on the subject, I think it would be neat to deploy one of their models to a videogame. I haven't really followed all the progress in reinforcement learning during the last year but I think a quick Google search will help you a lot.