Robot Hands and Multi-Fingered Haptic Interfaces

The following sections are included:

• About the Author
• Preface
• Contents
• List of Tables
• List of Figures

To enable a robotic hand to grasp a target object dexterously, like a human hand, structural features and control functions similar to those in human hands are necessary. Until now, a deep understanding of the human hand has been the province of medical professionals. Researchers in robotics have much to learn from them. This chapter describes the structure of the human hand, its grasping functions, and the structure of the humanoid robot hand called the Gifu Hand. Furthermore, the configuration of a multi-fingered haptic interface called the Haptic Interface Robot (HIRO), a new application for the robotic hand, will be discussed.

When a robot hand grasps and manipulates an object, the hand and object interact at several contact points. Since contact conditions vary according to point contact, line contact, surface contact, and the existence or non-existence of friction at the contact point, the constraint conditions that the object receives are also diverse. This chapter presents basic kinematics in terms of the grasp of a multi-fingered robotic hand, including contact models, forms of grasp, constraint conditions, and manipulable grasp. Introduction of these topics are available in Yoshikawa, 1995, 1996a, 1996b and Zeng et al., 1997.

When the robot manipulates an object, the kinematics of fingertips are constrained by the contact with the object, as presented in Chapter 2. The kinematics of an automobile are constrained because the front and rear wheels are only allowed to roll and spin, but not slide sideways. These motions are under the kinematic constraints. The kinematic constraints are generally classified into two types: holonomic constraints and non-holonomic constraints. The holonomic constraints are expressed by algebraic equations in generalized coordinates. The other constraints are non-holonomic constraints, which are typically expressed by differential equations in generalized coordinates. We know that independently driven two-wheel mobile robots can take any position and orientation on the plane in time. This is due to the effect of non-holonomic constraints…

This chapter reviews basic methodologies for creating dynamic models of robot hands. A dynamic model is a representation of the relationship between the joint torques and the dynamic motion of the robotic hand. The dynamic model is used widely in the simulation of robot motion, analysis of robot hand structures, and design of control algorithms. In this chapter, we will explore the dynamic model of a rigid body using the Newton–Euler method. Next, we will determine the dynamic model of a multi-joint link mechanism using the Lagrange method. When a robot hand with multiple fingers grasps an object, the robotic hand motion is restricted in order to maintain contact with the object. The dynamics model with constraint is derived.

Robotic hands are mechanical non-linear systems. This chapter outlines the basics of non-linear systems and Lyapunov's methods for stability analyses. Lyapunov theorem is useful in analyzing the stability of non-linear control systems. Proofs of theorems presented are outlined; more details may be found in the references at the back of the book (Slotine and Li, 1991; La Salle and Leffschtz, 1961; Bacciotti and Rosier, 2005; Imura, 2000; Queiros et al., 2000).

The cooperative control of multiple fingers on a robotic hand used to perform an operation by grasping a target object is a basic technique of dexterous manipulation. For many cooperative controls, widely used control laws are based on compliance control (Cutkosky and Kao, 1989), impedance control (Kosuge et al., 1995), control of underactuated fingers (Birglen et al., 2010), dynamic control (Zheng et al., 2000), together with position and force laws based on hybrid control (Hayati, 1986; Yoshikawa and Zheng, 1990). In these control laws, it is assumed that the dynamics of the system control is known. However, it is often difficult to find the exact dynamics of the system. In addition, the dynamic parameters of the system will vary depending on the object being manipulated. Cooperative control using a neural network (Hwang and Toda, 1994), learning control (Naniwa and Arimoto, 1995), control without object sensing (Ozawa et al., 2005), adaptive sliding control (Su and Stepanenco, 1995; Ueki et al., 2006) and adaptive coordinated control (Kawasaki et al., 2006) have been proposed to accommodate cases where the model parameters of the system are either unknown or fluctuating. This chapter describes impedance control for manipulating an object with multiple fingers, computed torque control, and adaptive control. Introduction of robot hand control may be found in the references at the back of the book (Mason and Salisbury, 1985).

Haptic interfaces are used to present force and tactile feeling to humans in cases of telecontrol for robots, manipulating objects in a virtual reality (VR) environment, playing games with force feeling, and so on. Some of these interfaces provide three-dimensional (3D) force feelings to the human fingertips, and a few haptic interfaces cover the workspace of the human arm. This chapter focuses on the control of a multi-fingered haptic interface in a VR environment, and on applications using the multi-fingered haptic interface.

Teleoperation of robots is often used in hazardous environments, such as outer space, the ocean floor, and nuclear power settings. A master–slave system is the type of teleoperation technology generally used, where the master is a haptic interface operated by a human, and the slave is a remotely located robot. The slave then moves according to motions generated through the master system. Human biological signals have also been used as a type of master system for the command of robots. One typical application is the myoelectric controlled hand. The electroencephalogram (EEG) and surface electromyogram (sEMG) have been used in many studies as a brain–machine Interface (BMI) because of their direct relationship to human motion. This chapter presents the fundamentals of the master–slave system, teleoperation of robot hands through a multi-fingered haptic interface, and prophetic hand control by sEMG.

Linear algebra is useful in the study of mechanical control systems in general, and in the state–space approach in particular. The subject of linear algebra is complex and is based on a rigorous mathematical foundation. In this appendix, the basics of vectors and matrices are reviewed.

The following sections are included:

• Chapter 1
• Chapter 2
• Chapter 3
• Chapter 4
• Chapter 5
• Chapter 6
• Chapter 7
• Chapter 8

The following sections are included:

• References
• Index