A Control Architecture for Efficient Point-to-Point Motions with Articulated Soft Arms

Introducing elasticity in the robot’s mechanics is proving to be a key ingredient in endowing robots with the ability of performing efficient and effective motions, especially when periodicity is involved in the task. However, how to develop controllers that can fully take advantage of these new possibilities is still an open challenge. This thesis tackles an instance of this general problem, by proposing a control architecture for executing efficient point-to-point periodic motions in robotic systems with adjustable parallel elasticity and variable stiffnesses. The algorithm is composed of two loops. The low-level one compensates for dissipation, therefore putting the robot in a stable oscillation. The high-level controller acts once every period. It adjusts the amplitude of excitation and the physical parameters of the system in order to move the end effector closer to the desired targets and, if possible, to compensate for any oscillation’s period error. Both loops are derived under a quasi-linear approximation of the dynamics, which allows for an analytic expression of the controllers. The resulting architectures show remarkable robustness both in simulations and experiments, across several conditions and tasks, combined with a high level of efficiency.