@squiddy

Bill of Materials updated(for a single leg) may still contain some mistakes.

Hi guys! This is my mini devlog starting to do more with the simulation

Change log:
Changed 3d model to be much more simplified to reduce computing power and get rid of texture issue from last time
Uploaded and ran the physics ragdoll in the simulation and added collision tags to the URDF so everything would collide with the ground
Use onshape-urdf tool to make more detailed changes to the joints, and add joint limits to the robot to maybe fix joint breakage.
Issues log:
For some reason the joints keep breaking in the simulation where it detaches from the original point after an impact into the ground for example

Hi guys! This is my full devlog starting to do interesting software

also can someone tell me how to edit the devlog with bold letters and paragraph spacing? I can do it now but only after posting and editing.

#7 Change log:

1. Researched and settled on Genesis World as the tool I will use for the physics sim as as well as RL training. Would use Issac Lab but it has poor support for AMD GPUs at the moment, so I decided to use Genesis World instead, which is a relatively new tool that doesn’t have as much documentation as other choices such as Mujoco. I setup everything in windows using a WSL according to official documentation, but pytorch was giving me a hard time and there was so many work arounds i had to do to accommodate for ROCM(amd gpu architecture) as well as windows that I decided I would just download linux and dual boot it on my computer, cleared a lot of headaches. It technically worked on windows but it would only use the cpu and wasn’t able to utilize GPU for parallel computing, leading to slower training times.
2. Set materials, joint limits etc to prepare for exporting as URDF using the tool Onshape to URDF
3. Imported URDF into Genesis world to test out the physics
4. Made a BOM will post soon

Issues log:

1. Yet to fix solder and fully finish soldering to a centralizedf power distribution
2. Finding name? please suggest
3. Need to fix URDF having no joints for some reason

thanks for reading pookies

Hi guys! This is my full devlog #6
Change log:

1. change from large branching power distribution to centralized power distribution due to the previous trunk and branch configuration being too prone to soldering breakage, Had to resolder joints multiple times was getting too annoying. 3D printed a xt60 power distribution central block.

2. Wrapped electrical tape and ziptied wire joints to alleviate stress

3. Made a crude wooden stand with scrap wood so its easier to work with the robot.Designed and 3d printed covers for the odrive motor controllers to protect the ports and solder joints

4. Tested 36V li-ion battery scavenged from my broken hoverboard and it works but no noticeable difference from 6s(22.2V) lipo, will stick with 6s lipo for now

5. Issue from last time where 4th motor wont work on CAN bus. Checked the CAN id’s for both axis(main and ghost) but all was correct. Must be a hardware issue related to the signal termination resistors. Fixed by removing two jumpers on the odrives that removed two 120ohm resistors from the parallel circuit, increasing resistance, so that the signal can successfully terminate.

Issues log:

1. Solder breaking off
2. Noticeable backlash on joints, didn’t think it was that large at first but when mounted on a long joint it is noticeable, hopefully won’t impact performance that much
3. Goals for next devlog:Inverse Kinematics test with leg in arduino IDE before figuring out motion reinforcement learning
4. Wire everything for one leg cleanlyFind a name for the robot for more aura tell me if you have ideas

mini devlog #5 lots of issues today:

1st; ran into issues with transceiver. when the program ran in arduino ide the esp32 just wasn’t able to command the motors to do anything, turns out I set the wrong baud rate and wasted an hour of my life. there also seems to be an issue with signal termination whenever I plug in a fourth motor into the CAN chain the program stops working from the esp 32(anyone with experience with this please help). this is the biggest problem I will need to solve.

2nd; I suck at soldering so all the joints started to bend and break off, I will need to design a better case for the xdrive mini boards so the solder of the xt60 connector doesn’t take the brunt of the flexing force. will need to re-solder most of these joints in order to fix the cold solder joints from before. at some point some magic smoke started coming out from one of the boards, no clue what its from but it solved itself and everything worked after re-plugging power supply. Should probably investigate further but i didn’t see any obvious shorts and im too lazy. Thanks for reading <3

Mini devlog #4 tested can transceiver and IMU unit, took a lot of troubleshooting connecting them to esp32 and using them on arduinoIDE, will switch to VS code soon

Mini devlog #3: 1. finished up cad, widened frame to accomodate for free rotation of the hip joint and added ziptie mount(temperory while find a better solution and figure out the wiring) added esp32 mount, imu mount, CAN bus transiever mount in progress, and added the odrive models in the cad(not shown in screenshot oops). 2. finished setting up motor firmware and tuned all actuators, found and issue with one of them having too much friction internally so I need to open it up and fix it(perhaps I printed the old version of the axle with the new design?). I soldered all the motors to the controllers and all the power connectors as well and tested them individually while in the cycloidal gearboxes. big thanks to justlovescience for his video on how to set up the brushless controllers, had to tweak quite a few things and the UI wasn’t working for me but was helpful regardless. mostly fixing mechanical bugs today, next steps are finalizing v1 of the whole robot and v3 of the leg, and upload model to the physics simulation. BOM coming soon. Thanks for reading allat

assembled both sides in CAD, debated on making a active ankle joint, but decided it would add too much weight and complexity since I would have to source a smaller motor. Keeps it simple to make all actuators the same. Working on electronics wiring and encasing soon

putting leg together in assembly