Modelling and Controlling Soft Interactions

Compliance is a key factor during physical interactions between agents and objects, allowing delicate and robust manipulation. Such a compliance can reside in human and robotic hands, as well as in objects present in the environment. This Thesis arose from the quest for mechanical conditions fostering object manipulation, and shaped as an investigation on techniques to model and control the compliance in soft interactions. Human and robotic hands are considered, functional and biomechanical models are discussed, devised and validated. To deal with unpredictable configurations assumed by intrinsically and passively-compliant underactuated robotic hands, a solution based on magnetic actuation is proposed. The exploitation of small magnetic elements allows also to simplify the robotic end-effector design by relying on local interactions to manipulate extremely deformable objects like garments. Moreover, mechanical forces exchanged during physical interactions are used to devise a control strategy for human-robot cooperative grasping, relying on a novel contact model exploiting linear elastic patches to ensure the contact permanence. The Research work contained in this Thesis shows that developing techniques for implementing robotic soft interactions is feasible and can significantly broaden the spectrum of robotic applications in real-world scenarios. An extensive experimental validation of the theoretic work supports the discussion.