Diverging results

Hi,

I am currently developing a control system for a 3 link manipulator as part of my project. I got some torque control values that is needed to be applied at each joint of the manipulator to produce desired trajectory. I've developed a matlab simulink model to simulate this as well(I dont want to use Sim mechanics model as of now), but the values produced by my model are diverging/blowing.

For instance even if I am providing holding torque (to counter-act gravity term, in this case angles of manipulator should not change), initial angle values do not change, but over a time period these values starts changing and eventually goes on to infinity.

Sorry for the long question.

Prompt response would be deeply appreciated.

[Jan]

"The initial values do not change, but... go to infinity"? I cannot follow. What exactly happens?

[Shobhit Gupta]

Hey,

I've shared the .mdl file, please go through it.

Answer 1

[Jan]

I guess, that you apply the counter-act gravity term with a limited accuracy. Then the small resulting torque will accumulate over the time.

There is no way to solve this drift in an open-loop control system, which acts on the torques only. To control the position of a robot, you have to control te position ;-)

[Shobhit Gupta]

Hey Jan, yes the counter-act gravity term is indeed little inaccurate (differs by e-13).

Are you sure this drift can't be removed? Basic robotic equations do have thetadot and thetadotdot (coriolis and acceleration terms in it as well).

The problem persists when I apply 0 torque to check if the system is okay(Ideally it should oscillate in a gravity field, with constant energy). My system is blowing up for any open loop torque I am providing to it. In that case, how can I check correctness of my system.

Thank You

[Arnaud Miege]

Can you upload your model to a filesharing site? There may be some modelling errors in the model of the robot.

[Shobhit Gupta]

Hey,

please find the files at this URL.

Simulink model file :

http://www.4shared.com/file/mzPgomP6/holdingtorque.html?

.mat file to give data into the modeled system.

http://www.4shared.com/file/YKF3Xzvt/torXv.html?

[Shobhit Gupta]

Hi,

It basically simulates torque = D(q)qdotdot+C(q,qdot)qdot+G.

If you'll run the model for holding torques(equivalent to gravity torque, it works by changing solvers) but it does not work for torXv which should produce a linear motion in the links. Please help!