• Model Based Design
• Robot Control
• Robotic Hands
• Soft Robots

The extended mind thesis states that cognitive processes in biological agents are not skull\--bounded, but they extend instead to the whole agent's world. In other words, intelligence is seen no more as the sole product of the brain, but as emergent characteristic of the interaction between body, brain, and environment. While the thesis is still debated in philosophy, the idea that intelligence can be outsourced to body and environment is a very appealing notion for roboticists.

Elasticity and hierarchical architectures are among the main drivers of this intelligence embodiment in animals. Tendon and muscles enable animals to robustly interact with the external world and to efficiently perform dynamic and oscillatory tasks. In this thesis, we propose strategies for reproducing these characteristics in artificial bodies through the selective introduction of impedance and under-actuation. In this way, we equip the robotic body with the ability of autonomously presenting intelligent behaviors, even if disconnected from the artificial brain.

In face of such a shift in perspective, classic control approaches reveal to not always be applicable or even suitable for producing the desired emergent behavior. As in the classic skull-bound view of cognition, these techniques recognize a predominant role to the controller which completely specifies a prescribed behavior for body and environment, washing out the intelligence introduced in design phase. In this thesis novel control approaches are proposed aiming at establishing more balanced interconnections between controller, robot’s mechanical structure, and environment.