LAB 3: INVERSE DYNAMICS JOINT CONTROL

FRICTION COMPENSATION

This lab focuses on implementing an inverse dynamics based control algorithm on the robot arm and then, compare it’s performance with a PD + feedforward controller. In order to implement an inverse dynamics based control law, it is important to model friction compensation torques well. The friction of each joint has been considered by implementing friction compensation using a Static friction model. As seen in the video, the motors supply the required friction torques without any control input and even a small jerk in any direction is sufficient to make the robot move freely as if it were in a frictionless environment.

TRACKING A CUBIC TRAJECTORY

In the above video, one can clearly see the inverse dynamics control law in action, making the robot arm follow a step input trajectory which is defined using cubic splines for all the 3 motor angles. The performance of this controller was comparable to a PD+feedforward controller but its advantages are clearly shown when the robot configuration is changed slightly. An example of this could be that if one puts an additional weight at the end effector, one would have to change all the gain values again for the PD+feedforward controller, whereas, for the inverse dynamics controller, only a change in the physical parameters is required without re-tuning the gains again.