Human-inspired software regulation of manipulator impedance for model-based compliance discrimination with soft optical sensors.

Compliance discrimination can play a crucial role in robotic manipulation and grasping. In previous studies, a model-based approach inspired by a computational model of human tactile perception was implemented with a soft biomimetic optical tactile sensor. By relying on optical flow computation, the rate of expansion of contact area under increasing probing forces was computed and used for compliance estimation. Additionally, by exploiting the observation that humans modulate muscular co-contraction to optimize compliance information uptake, it was proposed to extract human muscular patterns during softness probing actions and map them onto a one degree of freedom variable stiffness actuator that mimics the agonistic-antagonistic human muscle behavior. It was demonstrated that the model-based approach with the human-inspired stiffness regulation led to a high accuracy in compliance discrimination. In this thesis, I proposed to extend these results by mapping human muscular profiles to control a multi degrees of freedom manipulator, whose stiffness can be varied via software control.