A soft robotic approach for dexterous and robust grasping

Abstract

Nowadays, there is an increasing trend to develop robots for use in every corner of human daily life. Development of such robots demands a paradigm shift from main considerations of speed, power, and accuracy in indutrsial robots to primary concerns of safety, compliance, and adaptability in human-centered robots. Inspired by natural organisms, soft robotics has quickly aroused interests from the robotic research community. With inherent compliance, soft robots are an ideal artificial candidate to function safely and reliably like living creatures in human-centered environments. The rapid progress of material and fabrication technologies allow more sophisticated soft robots to be developed for a wide range of potential applications where robots are required to interact with humans, for instances, in grasping, manipulation, and delivery of objects to humans. In terms of grasping, lots of achievements have been reported based on the traditional rigid robotic approach which can perform grasping with high accuracy, high speed, and large power suitable for industrial applications. However, the structural complexity, size, control difficulty, and affordability are a trade off between the grasping performance and system reliability for rigid robotic hands. Even though highly desired for human-centered applications, the development of safe, compliant, and cost-effective grasping robotic hands is still a challenge using the rigid robotic approach. Soft robotic hands based on soft robotic technology have been shown to solve this grasping challenge effectively. Inherent compliance of soft robot hands provides excellent adaptability towards different objects with various geometrical and physical features. A successful grasping can be achieved by a soft robotic hand with both simple structural design and simple actuation. The continuum and compliant structure of soft robotic hands break the restriction of unitary morphology of rigid hands, joint link chains, which renders a soft robotic hand to realize bioinspired structure more effectively and to achieve effective grasping of novel objects. Thus, the soft robotic approach possesses promising potential to applications in daily human-centered environments. Although soft robotic hands present connatural priorities for solving grasping challenges that are difficult for traditional rigid robotic approach, yet there are four obvious deficiencies of soft robotic hands affecting their grasping reliability and limiting their wide spread use. They are the lack of in-hand manipulation dexterity, lack of grasping robustness, lack of sensory information, and difficulty to control. In this research, the main purpose is to address the above mentioned deficiencies of soft robotic hands and propose innovative methods to enhance the performance of soft robot hands. Methodology, modeling, design, and implementation of such methods are presented and illustrated in details. Prototype soft robotic hands and grasping control systems are built for dedicated experiments in order to validate the effective grasping performance. In particular, a fully actuated 26-DOF robotic hand with comparable human hand dexterity and in-hand manipulation capability has been developed together with its sophisticated control system. For robust grasping, a variable stiffness soft hand that can endure punching or piercing by objects with sharp spikes has been developed and deomonstrated in grasping tough objects such as durians, both in air and under water. A novel approach for self-position and force sensing in soft grasping is also elaborated with both theoretic modeling and experimental validation. This capability allows effective control on various soft hands. Overall, this thesis has presented a comprehensively study on novel approaches that have great potential for development of soft robotic hands with good performance of robustness and dexterity.