k-bot teardown

[Chris Albertson]

If you are a robot hardware geek, and I assume everyone here is, this first video is great.

“K-bot” is an new  open source humanoid robot.   It seems to be at a point of development now where it can stand on an uneven surface like grass and you can push it, and it takes steps to recover balance.  It can be teleoperated. The full hardware and software design is on GitHub.  You can buy one for about $10K or build it from the design files.

One design idea I like is the hands mount, exactly like a lens on a camera body.  They use parts from a micro 43 camera to make the mechanical and electrical connection.   The other design feature I like is what they call “clam-shell” but I’ve called it “exoskeleton” or “stressed skin”.  The part you see is not a cover but the actual structure.

But what I wanted to post is a hardware teardown.  Two designers do a complete teardown and take every screw and part off and tell you why they built it like they did.    They have design ideas that would work on other kinds of robots.   The design is astonishingly simple but strong and repairable.   

[Stephen Williams]

Nice!  I saw a reference, on a different project I think, using the micro 4/3rds lens mount.  Not a bad idea.  However, I could not find any reasonable source for both sides of the mating hardware.  If there is a good source, that is great.  If it includes the newer signaling connection block + pogo pins, way better.

Given laser cut metal parts, should be able to design 3D printed parts to create a similar solution.  And then it could be printed at different sizes, different strengths, with custom / robot-focused power + signal connections.  Need a plate that fits through another plate, locking upon twisting, with a center cylinder that aligns both.

The hand dock should not be at the wrist joint but further up the forearm, close to the elbow.  The forearm + wrist + hand should be a single unit.

The only major design deficiency I see is that I'm sold on having dual calf motors controlling the ankle joint for 2 degrees of freedom with differential control: Both motors go up, toe goes down.  Both motors go down, toe goes up.  If one motor is offset from the other, the foot swivels to cant left/right, at whatever angle both motors are set to.  I have pictures of several robots that use variations on this.

Two degrees of head control can be done with the same joint principle.  Then add rotation separately.

sdw

--
You received this message because you are subscribed to the Google Groups "HomeBrew Robotics Club" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hbrobotics+...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/hbrobotics/4E2FE396-6749-46C8-BD63-618A7ECADBD0%40gmail.com.

--

Stephen D. Williams Founder: VolksDroid, Blue Scholar Foundation 650-450-8649 | fax:703-995-0407 | s...@lg.net | https://VolksDroid.org | https://BlueScholar.org | https://sdw.st/in

[Alan Downing]

FYI, K-Scale was the startup that made the previously discussed open-source robot.  It's now bankrupt. Lots of details and lessons-learned in the following article:

https://eu.36kr.com/en/p/3559984314980485

To view this discussion visit https://groups.google.com/d/msgid/hbrobotics/82c0c21c-5015-4cde-bf1d-dc4092905f89%40lig.net.

[Dan]

Well here is lesson one. When a benefactor like me with 20 years experience emails the CEO of a company and offers their expertise and financial help like I did with Ben Bolte maybe they should have the courtesy to respond instead of the way kids today think that ghosting is OK.

To view this discussion visit https://groups.google.com/d/msgid/hbrobotics/CAAvYDnHNB07ZOruVF-XP_7oF6bh60Bbbua%3D0jYOh5vvVkYr%2BnQ%40mail.gmail.com.

[dpa]

Maybe not.  Perhaps the lesson is a simpler one: that humanoid robots, for all their appeal to investors, are actually not a good idea.

cheers!

dpa

[Chris Albertson]

On Dec 12, 2025, at 5:22 PM, Alan Downing <downi...@gmail.com> wrote:

FYI, K-Scale was the startup that made the previously discussed open-source robot.  It's now bankrupt. Lots of details and lessons-learned in the following article:

https://eu.36kr.com/en/p/3559984314980485

On Sat, Oct 18, 2025 at 2:04 AM 'Stephen Williams' via HomeBrew Robotics Club <hbrob...@googlegroups.com> wrote:

Nice!  I saw a reference, on a different project I think, using the micro 4/3rds lens mount.  Not a bad idea.  However, I could not find any reasonable source for both sides of the mating hardware.  If there is a good source, that is great.  If it includes the newer signaling connection block + pogo pins, way better.

Given laser cut metal parts, should be able to design 3D printed parts to create a similar solution.  And then it could be printed at different sizes, different strengths, with custom / robot-focused power + signal connections.  Need a plate that fits through another plate, locking upon twisting, with a center cylinder that aligns both.

The hand dock should not be at the wrist joint but further up the forearm, close to the elbow.  The forearm + wrist + hand should be a single unit.

The only major design deficiency I see is that I'm sold on having dual calf motors controlling the ankle joint for 2 degrees of freedom with differential control: Both motors go up, toe goes down.  Both motors go down, toe goes up.  If one motor is offset from the other, the foot swivels to cant left/right, at whatever angle both motors are set to.  I have pictures of several robots that use variations on this.

Two degrees of head control can be done with the same joint principle.  Then add rotation separately.

The trouble with differential drives in the ankles is the added mass.    The feet are the absolute worst place to add mass.   The joint needs to be fast and run on a fast control loop

Also, and this might be important, they simulate compliance or “spring” with motor torque and back drivable gear reductions.   The differential drive might be backdrivable, but I wonder if it can independently control torque in both axes?   Can you make one axis stiff and the other compliant?  I’d guess that you might need different “springs” on each axis.   And then “mass” is such a big deal with walking robots.

Those of us doing quadrupeds get around the ankle orientation problem by using spherical feet, rubber balls.   I did see one biped that also did this.  Actually, it was a quadruped where the legs had enough range of motion that the robot could walk on the front or rear pair of legs.  It was reasonably stable.

Baleria’s toe shoes are basically a point contact with the ground.  So we know that humans can run,  jump, and spin with point contact.   Flat feet just make it easier. 

A common compromise is to treat biped feet as if they were ice skates, with a line contact to the ground. Then we can ignore lateral bends in the ankles.