First prototypes for the motor rubber protection. Resin 3D print for silicone mold.
#k3lso #quad #quadruped #robotics #bldc #mjbots #MIT #cheetah #design #prototype #fusion360 #mold
1 Like
New design for low leg. A smoother translation to attach feet’s. Inspiration from #bostondynamics #spot
#k3lso #quad #quadruped #fusion360 #design #prototype #robotics #leggedrobotics #unitree #laikago #MIT #cheetah
2 Likes
Where did you manufacture these rubber protections?
I see you designed and printed samples for molds, but I am missing steps after that 
Same place as the alu and the mold plugs. 
1 Like
So you just sent them 3D models and they make everything?
Are the pictures of white silicone molds from you or from them?
More or less. 3D models, sometimes 2D drawings also.
The white silicone mold is also made by them. But yeah it cost bit extra but I save time.
1 Like
So for revision two on the lower legs I adding a smoother foot transition. The transition and the design itself is inspired by BD Spot.
My first design was a bit to long. So next version is 275mm foot included. This also saves a bit of torque.
When adjusting the size to ~ 275mm the Strength analysis also looks much better.
First prototype of the legs is all milled in ALU 7075. For second revison I might go down milling in CF. lets see.
1 Like
Excellent experience, but also looks like quite an investment (granted Spot Mini sells for $75K, there are “copies” from China in the <$20k range). Do you have a “day job”? If so, what do you do?
1 Like
I founded and co-founded a few different internet companies a few years back. Now I currently have my own small business as consulting with various stuff and yes and the moment one could say I have 3 different jobs.
1 Like
So I adjusted the length of the lower leg to 280mm. which makes upper / lower leg same length. 280/280mm. Modified the design back and forth and ran multiples FEM analysis to get the estimated strength correct. So revision two is more or less done.
I have also started to play around with some color choices (but haven’t decided on covers yet). But I’m leaning against same colors for covers as the rubber details.
The rubber details is Pantone cool gray 10c
2 Likes
@raess Your work is amazing! I love the attention to detail on all the parts.
I see you are using Gazebo. Your robot simulation seems to be working really well. I’ve been trying to get my humanoid simulation working and it’s been a struggle. I had trouble getting the surface and joint parameters like mu, damping, etc to work right. Made for some weird behavior with “soft floors”, crazy bounces and break-dancing robot until I used Hyperopt to auto-tune the parameters and this seemed to work good but still the robot is super jittery. Even if in the Gazebo plugin I am setting hard-coded SimJoint->SetPosition(0) the joints are jittery like epilepsy…you dont have this? If I comment out the SetPosition(0) then the robot isnt jittery. Strange.
I am involved in the latest versions of the Ros2_controls-Gazebo plugin so I am very familiar with the Gazebo API but like I said even hard coding those SetPosition calls causes jitter so it must be something in Gazebo or the URDF. I’d like to hear your experience. I am running Gazebo v11.1.
2 Likes
Thanks!
I am actually on pybullet atm. But yeah I will probably use gazebo later on again. Some problem I had with Gazebo back and forth on different models in the past.
I would start removing every single component until the jitterring has stopped (one by one)
If one or more components causing jitter and would double check and recalculate inertias for that component.
Maybe you can also disable contacts and collisions in gazebo until you find which of the links causing the jitter.
Not sure this helps. But it has helped be in the past.
1 Like
I just posted K3lso quadruped on
@hackadayio
will try to update this also as we go! 
https://hackaday.io/project/176487-k3lso-quadruped
Got on the Featured Projects almost instantly on http://Hackaday.io
#k3lso #quadruped #quad #robotics #ros #UPXtreme
1 Like
So you are thinking that a bad inertial tag might cause the problem? That is interesting. I got the inertials from solidworks and verified each origin/etc but I like your idea of removing parts of the model and see what happens. I am also using the inertial parameters in Orokos KDL dynamics calls and the force outputs from that seem sane. Any chance you could list your segment surface and joint damping parameters here? Or those parts of the URDF?
1 Like
