Innovative robot hand designs of reduced complexity for dexterous manipulation

Abstract

This thesis investigates the mechanical design of robot hands to sensibly reduce the system complexity in terms of the number of actuators and sensors, and control needs for performing grasping and in-hand manipulations of unknown objects.

Human hands are known to be the most complex, versatile, dexterous manipulators in nature, from being able to operate sophisticated surgery to carry out a wide variety of daily activity tasks (e.g. preparing food, changing cloths, playing instruments, to name some). However, the understanding of why human hands can perform such fascinating tasks still eludes complete comprehension.

Since at least the end of the sixteenth century, scientists and engineers have tried to match the sensory and motor functions of the human hand. As a result, many contemporary humanoid and anthropomorphic robot hands have been developed to closely replicate the appearance and dexterity of human hands, in many cases using sophisticated designs that integrate multiple sensors and actuators---which make them prone to error and difficult to operate and control, particularly under uncertainty.

In recent years, several simplification approaches and solutions have been proposed to develop more effective and reliable dexterous robot hands. These techniques, which have been based on using underactuated mechanical designs, kinematic synergies, or compliant materials, to name some, have opened up new ways to integrate hardware enhancements to facilitate grasping and dexterous manipulation control and improve reliability and robustness.

Following this line of thought, this thesis studies four robot hand hardware aspects for enhancing grasping and manipulation, with a particular focus on dexterous in-hand manipulation. Namely: i) the use of passive soft fingertips; ii) the use of rigid and soft active surfaces in robot fingers; iii) the use of robot hand topologies to create particular in-hand manipulation trajectories; and iv) the decoupling of grasping and in-hand manipulation by introducing a reconfigurable palm.

In summary, the findings from this thesis provide important notions for understanding the significance of mechanical and hardware elements in the performance and control of human manipulation. These findings show great potential in developing robust, easily programmable, and economically viable robot hands capable of performing dexterous manipulations under uncertainty, while exhibiting a valuable subset of functions of the human hand.