Motion planning - joint control inaccuracy

[Alsonax]

Hi everybody,

I'm trying to use V-REP to simulate a 6DoF robot arm. The aim to grab and reposition small objects on a table. As the gripper, I've copied the one from youBot which works fine. I'm using code frm motionPlanningAndGraspingDemo to generate a path to a target and then execute the motion with joints in torque/force mode.

The problem is that unless I use the Vortex dynamics engine, the motion is very inaccurate because of arm being very springy and leaning down because of weight.

I've tried adjusting the link weights and joint parameters, without success.

Is there a way to make it as stable as the demo scene using the Bullet/ODE engine?

Scene: https://www.dropbox.com/s/gvxlxwaeoerlp ... n.ttt?dl=0

Thanks a lot

[coppelia]

Hello,

• increase the inertia of your links by a factor 4 at least.
• try to set the P parameter of the PID controller 0.5.
• The object Rectangle7 is static, but built on a dynamic structure: that will lead to strange behaviours. Make it non-static, and attach it to its parent via a force/torque sensor
• Set a realistic upper velocity limit for your joints (e.g. not much higher than 180 degrees/sec
• Your manipulator is a heavy industrial robot: it is not meant to collide with its environment. You could simply make it all static, except for the gripper, in order to be still able to grasp objects.

By doing above, you will get a very stable simulation also with the other physics engines.

Cheers

[Alsonax]

That did it,

Thanks!