# Torque control in eye-in-hand visual servoing

In most papers about IBVS the camera velocity is computed and then used as a pseudo-input for the manipulator. (e.g. this one) Is there any work in which the dynamic Lagrange model $H(q) \ddot q +C(q,\dot q)\dot q+g(q)=\tau$ of the manipulator is taken into consideration in order to compute the torque required to move the joints accordingly?

• When you say, "dynamic Lagrange," do you mean the open-loop dynamics? I'm not sure the Lagrange dynamics apply to a controlled system, especially if you've used a state feedback controller to move the system poles. Nov 3 '15 at 2:04
• I am not sure why would you mix dynamics with kinematics. In most cases the torque controller (cascaded with the other 2) is used to seperate you from the dynamics and allow a kinematics based controll (you only care for positions velocities and accelerations, and forget about the forces and torques). From a different point of view, you can find lagrange model that computes the torques for you and add the kinematics on top of that and you are done
– 50k4
Nov 3 '15 at 8:31
• @Chuck I have added the equation I was refering to. The first step of the control algorithm just computes the velocity the end-effector/camera must have in order for the features to reach their respective targets. But in a real situation it would be best to compute the torque required to move the joints of the manipulator accordingly. Nov 12 '15 at 7:04
• Isn't that just simply the computed torque method? Nov 12 '15 at 7:44