• Modeling Control and Learning for Soft Robots
• Impedance Control and Regulation
• Flexible Robots
• Natural Machine Motion

Traditionally, robots were designed to be bulky and heavy to obtain good positioning performance. Due to this nature, they have been confined to industries. Indeed, their cumbersome structure may lead to dangerous interactions with humans or unstructured environments. The introduction of embedded elastic elements gave robots the ability to safely cooperate with people. Thanks to this evolution, robots held the promise of venturing out from factory lines. However, the advent of these technologies brought also the need of new planning and control methods. One of the most significant examples is related to the classic control algorithms. If used naively, high-gain feedback controllers would alter the desired compliant behavior of soft robots, hindering its advantages. Moreover, different tasks may require different mechanical behaviors to be achieved. So, planning for a system whose compliance may vary is a problem which is inextricably intertwined with defining trajectory and interaction plans. These two problems of great relevance to soft robotics are tackled in this thesis by developing new theoretical and practical instruments. These tools can be used to define plans and controls able to preserve the natural behavior of soft robots. These novel methods have been tested on variable stiffness robots, impedance-controlled manipulators and on a series elastic quadruped. Finally, I exploit these concepts to develop a new articulated soft robot for pick-and-place tasks.