@PitaGriffin @Mankaran32 ;)

@andrewcyu @Mankaran32 What do you mean by round of sim2real ? The goal is to learn to walk in simulation and transfer it on the real robot

@Mankaran32 Checkout the project here : The v2 was co-designed with Augustin Crampette !

@euluisfilipe @X0Radi I haven't started working on the v2, waiting for a very good walk on the v1! My plan is globally to make it cheaper and better mechanically designed so that it's more robust and easier to assemble.

@DrTimBlakely Check this out :)

@X0Radi It's not for sale, but you can build one ! I would recommend waiting for the v2 however, this design has some issues

@praveenanasurya @LeRobotHF's bio should have a working invite link

@Neurotronic67 Great ! What's your issue with the hip roll spacing ? What motors do you plan on using ?

@KyleMorgenstein @roboticseabass Here are some plots with all the observation inputs to the policy. On the left is mujoco, on the right is the real robot. The last 15 entries are the previous action outputted by the policy

@roboticseabass @KyleMorgenstein But right now, when running the policy on the real robot, I have a shaking problem. I'm investigating it !

@KyleMorgenstein Right now I have a pretty good walk in sim, that transfers well from Isaac gym to Mujoco, but not to the real robot yet. Sim2real is hard :)

@KyleMorgenstein I don't do anything specific for collision meshes. It's my understanding that Isaac gym makes convex Hull approximations automatically if collision meshes are not provided. Be my guest! You can find the URDF here

RT @planeterobots: @pollenrobotics annonce officiellement aujourd'hui le lancement de la nouvelle version de son robot humanoïde Reachy 2,…

RT @pollenrobotics: Reachy 2 is live! 🤖 Versatile, open-source, and ready for advanced tasks. Push robotics forward with us! Discover more…

@Neurotronic67 @Thom_Wolf Great! Let me know if you need any help :)

@Neurotronic67 @Thom_Wolf But they are very expensive unfortunately