Biomechatronic Design and Control of an Anthropomorphic Artificial Hand for Prosthetic and Robotic Applications
Abstract
This paper proposes a biomechatronic approach to the design of an anthropomorphic artificial hand able to mimic the natural motion of the human fingers. The hand is conceived to be applied to prosthetics as well as to humanoid and personal robotics; hence, anthropomorphism is a fundamental requirement to be addressed both in the physical aspect and in the functional behavior. In this paper, a biomechatronic approach is addressed to harmonize the mechanical design of the anthropomorphic artificial hand with the design of the hand control system. More in detail, this paper focuses on the control system of the hand and on the optimization of the hand design in order to obtain a humanlike kinematics and dynamics. By evaluating the simulated hand performance, the mechanical design is iteratively refined. The mechanical structure and the ratio between number of actuators and number of degrees of freedom (DOFs) have been optimized in order to cope with the strict size and weight constraints that are typical of application of artificial hands to prosthetics and humanoid robotics. The proposed hand has a kinematic structure similar to the natural hand featuring three articulated fingers (thumb, index, and middle finger with 3 DOF for each finger and 1 DOF for the abduction/adduction of the thumb) driven by four dc motors. A special underactuated transmission has been designed that allows keeping the number of motors as low as possible while achieving a self-adaptive grasp, as a result of the passive compliance of the distal DOF of the fingers. A proper hand control scheme has been designed and implemented for the study and optimization of hand motor performance in order to achieve a human-like motor behavior. To this aim, available data on motion of the human fingers are collected from the neuroscience literature in order to derive a reference input for the control. Simulation trials and computeraided design (CAD) mechanical tools are used to obtain a finger model including its dynamics. Also the closed-loop control system is simulated in order to study the effect of iterative mechanical redesign and to define the final set of mechanical parameters for the hand optimization. Results of the experimental tests carried out for validating the model of the robotic finger, and details on the process of integrated refinement and optimization of the mechanical structure and of the hand motor control scheme are extensively reported in the paper. Index Terms—Biomechatronic design, biorobotics, hand motor
I. INTRODUCTION
THE HUMAN hand represents a wonderful example of a natural biomechatronic system, which still represents a benchmark for robotic designers aimed at replicating its complex functionality [1]–[3]. In the literature, several examples of robotic hands can be traced, ranging from simple grippers for industrial applications up to more sophisticated artefacts trying to mimic human mechanics [4]–[7]. Typically, end effectors of industrial robots are simple grippers or specific tools able to perform stable grasp of a limited set of known objects. They are purposively designed for a specific task, showing high dexterity in task-oriented preprogrammed applications in structured scenarios, but featuring low anthropomorphism and low manipulation capability. Humanoid robotics is one of the fields currently devoting the most significant efforts to the design of artificial anthropomorphic hands. This is because humanoid robots are expected to achieve performance as close as possible to humans, trying to replicate them from the viewpoint of sensori-motor coordination, as well as of prompt reaction and adaptation to dynamic unstructured environments. Prostheticswas one of the first application fields envisaged for artificial anthropomorphic hands, for obvious aesthetic as well as functional reasons. Prosthetic applications of robotic technologies impose a series of challenging requirements regarding the cosmetic appearance, the size and theweight of the hand, and its embeddable control system, which is crucial for obtaining reliable and robust hand acceptable for end users. Commercial prosthetic hands are basically simple grippers with few degrees of freedom (DOFs), a limited biomorphic appearance, and one actuator able to exert high grasping forces [8]. Consequently, the control is usually very simple but robust; a couple of commands sent by the user control gripper opening and closure. Research is addressing control algorithms [9], [10] and some of them are based on neural approaches [11], i.e., the control action is often taken as proportional to the superficial electro myographic (EMG) signals extracted by surface electrodes applied to a couple of antagonistic user’s residual muscles. The control of the Ottobock prosthesis is paradigmatic for a typical prosthetic control [11]. In [10] and [12], a hybrid control is presented, where a digital controller operated by means of myoelectric signals converts the user’s grasping intention (as specified by the EMG signal) into an order for the control of the prosthesis. In it, sensors to the prosthesis and feedback to the user have been added [10], [13]. Anyhow, present commercial prosthetic hands still have a series of limitations

Download full report
http://googleurl?sa=t&source=web&cd=2&ve...291586.pdf%3Farnumber%3D4291586&ei=fgE9Tu7CN4ftrQf14uAZ&usg=AFQjCNGxVwljXdheQLYSxTzRU49CJGkeuQ