Chosing appropriate solver in simulink for robot simulations? Fixed step/Variable step?

[s.sa...@gmail.com]

Hello all,

Does anyone know what would be ideal solver for robot simulations? Fixed step / variable step / explicit / implicit / discrete / continuous?

I am working in Simulink. My work is focused on robot control, so I am trying to develop different control algorithms (Inverse dynamics, Robust, Adaptive..) using robotics toolbox as much as I can. So far I have just used fixed step, because if I chose a variable step the simulation seems not to move from 0% for more that 10 minutes, so I stop it every time. I guess that I am not setting something correctly.

If anyone knows the answer or if anyone could tell me of some literature where the pros and cons for those different solvers are written related to robot simulations it would be very helpful.

Thanks in advance,

Satja

[Peter Corke]

It depends a lot on the non-linearities in your model.  Very harsh non-linearities like Coulomb friction really kill a variable step solver.  Fixed step deals very easily with this, but the answer is less precise, gets more precise as step size gets smaller, but then that takes longer.

Perhaps remove any Coulomb friction to get going.  The mdl_puma560 has such friction.

peter

[s.sa...@gmail.com]

Thanks Peter. I have designed my own robot model based on puma 560 from toolbox, and I am using coulomb friction. I know now what kills the variable solver.

Thanks