Bio-inspired design of sensors and mechanisms in humanoid robotics

Many properties of the human hand contribute to its versatility for manipulation task, such as its complex sensory apparatus, its intricate anatomy and muscular system. Yet, neither the exquisite tactile human sensibility, nor the intricate anatomy or the muscles actuating the hand and the wrist can explain the manipulation skills of normal humans. Fine manipulation tasks are possible thanks to automatic control processes that manage the flow of sensory information and correlate it to the right motor commands. Currently, a state-of–the-art artificial hand can be compared to an almost totally anesthetized human hand. Such poor sensorization makes even the simplest manipulation task difficult to control. This PhD Thesis presents a novel bio-inspired approach to the design of artificial hands with a smart sensory system already embedded. The development of an advanced humanoid hand with a high number of DoFs is useless without the design of an adequate and reliable set of proprioceptive and exteroceptive sensors. Furthermore, another milestone in this integrated design strategy is the awareness that the output signals of all sensors have to be complementary. The result of their fusion into the control architecture has been considered in advance to develop a smart and effective local low level control for grasping and dextrous manipulation of objects. In this work, the novel artificial hands and sensors developed are presented, together with original integrated low-level control modules for grasping and detecting softness and roughness of objects.