Robust Grasp with Compliant Multi-Fingered Hand

As robots find more and more applications in unstructured environments, the need for grippers able to grasp and manipulate a large variety of objects has brought consistent attention to the use of multi-fingered hands. The hardware development and the control of these devices have become one of the most active research subjects in the field of grasping and dexterous manipulation. Despite a large number of publications on grasp planning, grasping frameworks that strongly depend on information collected by touching the object are getting attention only in recent years. The objective of this thesis focuses on the development of a controller for a robotic system composed of a 7-dof collaborative arm + a 16-dof torque-controlled multi-fingered hand to successfully and robustly grasp various objects. The robustness of the grasp is increased through active interaction between the object and the arm/hand robotic system. Algorithms that rely on the kinematic model of the arm/hand system and its compliance characteristics are proposed and tested on real grasping applications. The obtained results underline the importance of taking advantage of information from hand-object contacts, which is necessary to achieve human-like abilities in grasping tasks.