You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I am currently working on a project using PyBullet as a robot simulator and Baxter as the studied robot.
When starting, I used the demo showed in this repo. However, I observed strange behaviors in collision management and overall dynamics.
These behaviors are explained by the fact that in the accurateIK() method, joints are reset at every iteration of the inverse kinematics computation. Therefore, the resulting behavior is completely false.
It would be interesting to modify this in order to help future users of this demo.
Thank you !
The text was updated successfully, but these errors were encountered:
Hello,
I am currently working on a project using PyBullet as a robot simulator and Baxter as the studied robot.
When starting, I used the demo showed in this repo. However, I observed strange behaviors in collision management and overall dynamics.
These behaviors are explained by the fact that in the accurateIK() method, joints are reset at every iteration of the inverse kinematics computation. Therefore, the resulting behavior is completely false.
It would be interesting to modify this in order to help future users of this demo.
Thank you !
The text was updated successfully, but these errors were encountered: