Human interaction control with redundant robots

Collaborative robotics is an expanding research topic from both an academic and industrial point of view. In fact, there exists a high number of applications that it is not possible or convenient to completely automatize and therefore require the activity of a human operator. At the same time, it is extremely important that such activities do not cause physical or mental fatigue to the worker. In this context, the presence of robots that can safely interact and cooperate with humans is crucial. This thesis addresses the theme of collaborative robots, focusing in particular on a hand guiding application with a redundant Franka Emika Panda robot, studying the best control strategy for this framework. The analysis has two main parts, namely end effector control and null space control. As for end effector control, the objective is that the human operator is able to drive the end effector along a desired path in a precise, smooth and effortless way. Impedance control schemes provide poor results, and this is shown to be due to the shortcomings of the torque interface of the Franka robot. Another strategy is tested, called ``helping'' controller, which sends to the joints control torques aimed at minimizing the force that the user needs to provide to guide the end effector. This scheme has better results, but it is not suitable if also null space commands are sent. Instead, admittance control, gives the best results, also being compatible with null space commands. As for null space control, two innovative strategies are introduced, finding the null space commands as solutions of constrained optimization. These techniques are compared and tested in a simplified case study on the real robotic platform to show that their computing times are compatible with the real time requirements of the Franka robot.