# Implementing a torque-controlled method on a position-controlled robot

I am working with a position-controlled manipulator. However, I want to implement a torque-controlled method on this robot. Is there any way to convert a torque command to a position command?

I try to find research papers on this but I have no idea where I should start or which keywords I should use in searching. Do you have any suggestion?

• Can you describe the problem in more detail? You might want to look into 'impedance control' technique. – Eugene Sh. Dec 1 '14 at 18:32
• @EugeneSh. Actually impedance control is what I want to implement on the robot. The problem is that I can calculate a set of input joint-torque commands but the robot is joint-position controlled. So I would like to know if there is any way to translate joint-torque commands to joint-position commands. – Petch Puttichai Dec 2 '14 at 3:15
• Use PID controller. Your input will be the desired position. PID controller generates a torque that moves the manipulator to the desired position. – CroCo Dec 2 '14 at 18:40
• @PetchPuttichai But what the arm should actually do? Follow some trajectory? Point to point movement? Movement with maximum exerted external force? If you need a torque/force or impedance control, you will need torque/force feedback from sensor or from observer based on your motors feedback. Are you able to provide it? – Eugene Sh. Dec 3 '14 at 15:08
• @EugeneSh. The task may be to assemble furniture parts, or something like that. I think I have to implement impedance control in some way but now I still do not have concrete understanding or clear picture about it. Feedback information is available. – Petch Puttichai Dec 6 '14 at 9:11

We find this recent paper by Andrea Del Prete, Nicolas Mansard, Oscar Efrain Ramos Ponce, Olivier Stasse, Francesco Nori quite interesting:

Implementing Torque Control with High-Ratio Gear Boxes and without Joint-Torque Sensors

The authors presented a framework for implementing joint-torque control on position controlled robots.

Robotic arm are basically controlled using torque input to reach a position reference. Most common control techniques for such robot are computed torque (also called inverse dynamics approach), and P-D controller with gravity compensation.

You can find all the information you need on the bible of robot modeling and control, Robot Modeling and Control, Spong W. and al., 2005.

PS: searching papers about that subject is a very bad idea since this is no new controller that you are looking for, this is well established control techniques.

Many robotics textbooks should have descriptions on control algorithms for arm. Typically, you will wrap position or velocity control around torque control. You can use PID control for this, but i think there are other methods.

• This is not what OP asked. He is interested in torque loop wrapped around position loop. – Eugene Sh. Dec 3 '14 at 14:59

Check out this paper by Oussama Khatib, Peter Thaulad, Taizo Yoshikawa and Jaeheung Park from the Stanford Artificial Intelligence Laboratory: