Issues Log

2/19/19:

  • Change the mounting method and position of NUC, might interfere with manipulator’s aluminum rod, move the controller board if possible
  • Order smaller batteries to fit in 1st layer, return the bad battery
  • Resolve issue that boot up order may affect the functionality of OpenCR controller board
  • Configure Jetson TX2 (higher priority)
  • Order M2.5 long screws (~25mm), double-check length later
  • Camera mounting
  • Choose IMU model
  • Resolve issue that rosserial requires root privilege
  • All-in-one roslaunch file and alias in ~/.bashrc for easier use
  • Table clean-up, find lost items, label boxes

3/26/19

  • We had major issues with some of the Dynamixel servo-motors. Servo-motor at wrist replaced. It was a Dynamixel MX-106T and we are now using a MX-28 with a 25.5 Kg.cm stall torque which is sufficient for our task and weighs only 72g, versus 153g. for the MX-106T. It is also lighter and lower cost than the previous one. We had to make an adaptor to be able to re-use the same tray as before.
  • The shoulder bracket had to be extended, so that the shoulder joint can close further without collision with the servo-motor itself, and so bringing the tray much closer to the mobile base for a smaller footprint

4/4/2019

  1. Resolve the lack of traction problem by moving the manipulator to the back
  2. Set up Jetson TX2 and find a way to mount on the robot
  3. PCB mounting plate
  4. IMU mounting plate
  5. Re-wire electronics using the power distribution board
  6. Calibrate dynamixels, 3 different configurations for moving, grasping and dumping
  7. Resolve the wrist motor software limit
  8. Experiment on by-passing OpenCR board, try controlling chassis motor together with joints

4/11/2019


Got good hyperparameter for vision pipeline and did some reordering to make the speed even faster. Reduced False positives from 10-12 to 1-3.
Made program to track alvar tags in realtime and output their poses with respect to the robot frame.

09/11/2019

Create a software interface architecture. It’s a very first draft and the exact input/output are not defined, but just so we can start the conversation. Feel free to add comments or make a new version in a new tab! (please review @Laavanye Bahl @Changsheng Shen @Nithin Meganathan )

https://www.draw.io/?state={“folderId”:”13_zyzJ4dwR1OoO1rqVsqjA7-PAk-XVsw”,”action”:”create”,”userId”:”100643945105721275418″}#G1vzYdot8favcvLAjIbhXLW4opkA-yx-PT

I summarized our conversation of what our strategy for picking and mapping is here – I also added some math to start the conversation of the timing analysis. I estimate it will take us 23.2 min to pick up 15 toys. The biggest difference is picking up two toys at once for smaller toys(can save 10 min) and speeding up picking up and dropping toys (4 min each). (please review @Laavanye Bahl @Changsheng Shen @Nithin Meganathan )

https://docs.google.com/document/d/1uIMIBEK-43NzEoGURRYresxmMonYisqX67hMjqOMOqo/edit

10/26/2019

Overshoot in position and orientation, tune controller gains.

Need to get higher odometry update rate.

Integrate the vision pipeline with mobility and control.

Validate pick-up.

Dropping and box alignment need to be implemented and tested.