Zima · 2025-12-08 · 6 hours
Goal: Polish and test arm kinematics on hardware
What I did:
- Added joint and workspace limits
- Further tested ik solver in simulation
- Completed ros2 connected arm controller
- Sucessfully Tested ik on hardware
- Debugged firmware issues
- needed to downgrade esp32 board library to 2.0.x
- Found bug in the jacobian estimation code
What worked:
- 
- Teleoperated sample pick and place tasks (eg. pick up and stack rubiks cube)
- fixing jacobian estimation bug improved stability a little, but still lots of jumps in the solution space
- need to add penalty term for large changes in the joint state
What failed:
- IK solver good in simulation and fast, but certain awkward angles in real life cause jittery movement
Key learning:
- IK solver on its own can be a little unstable
- Good idea to have ai proofread your code
Next session:
- Add penalty for large joint state changes in the error calculation
- Test IK in real life again
- Begin writing data collection code for more complex demos
- arm + nav + 2 camera data
- finish and test multi camera driver
- visualize arm side by side with real life to see modelling accuracy
Time spent: 6 hours