Underactuated Robotic Hands

Underactuation is a widely used and a relatively old concept in robotics. Basically, it expresses the property of a system to have an input vector of smaller dimension than the output vector. Practically, in robotics, it means having fewer actuators than degrees of freedom (DOF). Applying this concept to robotic grasping arises from a simple fact: it is desirable to be able to grasp objects using a simple control rather than having to command and coordinate several actions. The idea behind underactuation in grasping is to use an ingenious mechanical system that can adapt to the shape of the object automatically. This mechanical intelligence, embedded in the hand, is based on the principle of differential systems. The latter devices automatically distribute one input to several outputs, the ratio between the different outputs being determined by the design parameters and the output states themselves. The same philosophy of intelligent design is commonly found in mechanical linkages where the different link lengths and joint types are determined at the design stage in order to follow a particular trajectory. If this trajectory is entirely predetermined, then only one DOF suffices to follow it. Yet, if this trajectory is too complex, a multi-DOF mechanism capable of following any trajectory is often chosen. Similarly, in robotic grasping, anthropomorphic hands have been predominant for years in research laboratories while industrial needs were often satisfied by a few simple grippers. The postulate behind the use of underactuation in grasping can be stated as follows: if the task to be performed is grasping, it should be possible to accomplish this one action using one single actuator.

Two main approaches dominate the literature on robotic grasping, namely, on one hand purely theoretical work on grasping and manipulation and, on the other hand, the rather intuitive design of functional prototypes. This chapter attempts to bridge this gap for the special case of underactuated fingers. Indeed, although the development of underactuated fingers aims at overcoming the theoretical difficulties of general manipulation issues and at obtaining prototypes of practical relevance, the capabilities of these fingers remain not well known. Prototypes have often been built through intuitive design, without a generic knowledge of the resulting behaviour and based mainly on special purpose computer-aided simulation. This chapter presents an effort to establish a common framework using simple theoretical bases to analyze the contact forces generated by robotic fingers during enveloping grasps. The fundamental goal of underactuation being simplicity, the objective of this work is to provide practical tools for the analysis and comparison of underactuated fingers. Indeed, some issues have been overlooked in previous work and should be systematically addressed. For instance, the grasp force distribution, the capability of the finger to actually exert forces on a grasped object, the stability of the grasp and others will be covered in this chapter. Underactuation in robotic hands generates intriguing properties, e.g. underactuated hands cannot always ensure full whole-hand grasping. Indeed, the distribution of the forces between the different phalanges is governed by the mechanical design of the hand since only one actuator is used and some phalanges may not be able to actually exert any effort in certain configurations. This uncontrollable force distribution can also lead to unstable grasps: a continuous closing motion of the actuator tending to eject the object, as discussed in more details in Chapter 4. A new method to study the capabilities of underactuated fingers is presented that allows rigorous comparison of different transmission mechanisms through the definition of indices that are similar to the dexterity in kinematics. In this chapter: The first part of the chapter (Section 3.2) establishes the fundamental background of our analysis and requires knowledge in screw theory and mechanical transmission design. It is then demonstrated in the second part (Section 3.6), how these results can be used to characterize underactuated fingers. Throughout the first part of this chapter, linkage-driven fingers using a mechanical architecture similar to the SARAH hands are used as an example but the methodology is general and other transmission techniques are presented in Section 3.6. Also, the methodology used in the subsequent sections to develop a general static model of underactuated fingers remains valid, even if the finger is fully actuated. Hence, the title of this chapter does not contain the word “underactuated.”

If one or more phalanx force is negative when an underactuated finger closes on an object, the corresponding phalanx will loose contact with the latter object. Then, another step in the grasping process will take place: the remaining phalanges—corresponding to positive forces—will slide on the surface of the object. This sliding process will continue until either a stable configuration is achieved (with only positive or zero phalanx forces) or joint limits are reached (usually a stable situation, but the shape adaptation is less effective) or the finger will curl away and loose contact with the object (ejection).

This chapter presents the optimal design of underactuated fingers considering several issues among which the force isotropy of the grasp, i.e., its ability to generate a uniform pressure on the object seized, its stability, and the Cartesian directions of contact forces. First, the force isotropy property is defined and a method to achieve the latter is presented. Its robustness with respect to undesired variations of the design parameters is also discussed. Second, guidelines to minimize or completely avoid ejection are presented as well as their influence on the desired optimality. The occurrence of the situations where one phalanx force is negative should be decreased (prevention is impossible as established in Section 4.2.7, at least with our model of underactuated fingers) and the ejection should be avoided.

Common robotic hands do not usually comprise one single finger, except maybe in tentacle inspired systems. The prototypes presented in Table 2.1 have a number of fingers comprised between two and five, while over 50% have three fingers. It is therefore a natural step to extend the principle of underactuation to the hand itself in addition to individual fingers. The purpose of the underactuation between the fingers is to use the power of one actuator to drive the open/close motion of all the fingers of a robotic hand collectively. The transmission mechanism must be adaptive, i.e., when one or more fingers are blocked, the remaining finger(s) should continue to move. When all the fingers are blocked, the force should be well distributed among the fingers and it should be possible to apply large grasping forces while maintaining a stable grasp. Introducing underactuation between the fingers of a robotic hand allows to further reduce the complexity of the systems, from the actuation point of view.

Following the analyses and optimizations discussed in the preceding chapters, practical aspects are now discussed in the context of design and control of underactuated hands. Several prototypes of underactuated hands built at Université Laval are extensively presented in this chapter. The analysis and optimization of underactuated fingers (Chapters 3–5) as well as the available techniques to extend the underactuation between the fingers (Chapter 6), naturally lead to the design and control of actual prototypes. Such prototypes demonstrate the behaviour and capabilities of this technology. First and foremost, in an underactuated hand, several fingers are required to grasp an arbitrary object. Hence, the first section of this chapter focuses on selecting the necessary number of fingers and their relative positions. The hands presented in this chapter are mainly intended for use in hazardous environments, e.g. in space. Successful grasping in a spatial environment is an important issue addressed in the literature. More specifically, in (Foster and Akin 2001), a study of the different grasping requirements for 242 existing crew aids and tools as well as during extravehicular activities (EVA) showed that over 50% of the grasps were cylindrical and three-fingered hands achieve over 90% of the required tasks. The results presented in the latter reference were used to orient the design choices presented in the subsequent sections.

This book is an attempt to lay the fundamental bases for the use of underactuation in grasping robotic hands and to summarize more than a decade of research. It covers a broad spectrum ranging from the analysis of theoretical properties of underactuation in grasping to the actual design and control of prototypes. In the introduction of the book, a survey of current architectures of robotic hands was presented. Then, the principle of underactuation, i.e., using fewer actuators than DOFs, was presented as well as how it can be used to drive robotic hands. Several prototypes, some of them century-old, were pointed out from the literature. Although underactuation is a very old technique, it has only very recently attracted the attention of the research community. Beside this introductory survey, the first contribution of the book is the definition of two matrices—termed Jacobian and Transmission matrices—allowing immediate characterization of the force capabilities of a robotic finger. These matrices provide the analytical expressions of contact forces developed by a finger of arbitrary design and with any number of phalanges. The only limitation of the method presented is that abduction/adduction actuation should not participate in the grasp effort. Analyzing these matrices and the resulting contact forces expressions lead to considerations on equilibrium, and can also provide practical tools for the designer to compare between different solutions.