Design and Control of Artificial Robotic Hand

Ali Bin Junaid*, Muhammad Raheel Afzal*, Tahir Rasheed**, Sanan Tahir**, Sharjeel Ahmed**, Mehreen Sohail** and Muzaffar Ali**

*School of Mechanical & Aerospace Engineering and ReCAPT

Gyeongsang National University, Jinju, , 660-701, Republic of Korea {alibinj & raheel379}@gmail.com

**Department of Mechatronics Engineering Air University, Islamabad 44000, Islamic Republic of Pakistan

{090434,090437,090483,090436}@students.au.edu.pk muzaffar.ali@mail.au.edu.pk

Abstract – Artificial robotic hands are designed having dexterity and functionality as close as the natural human hand produce.This paper proposes a bio-mechatronic approach for the design and control of an anthropomorphic artificial hand capable of performing basic human hand motions with fundamental gripping functionality. The mathematical model is derived using forward kinematics and also simulated on MATLAB® to ascertain the position of robotic fingers in 3D space. The dexterity of the artificial hand is exhibited by imitating the natural motion of the human fingers. Imitation is achieved by two different methods; a camera based marker recognition system to identify the human hand gestures and acquired flexion data from sensors attached to the human fingers. In order to have proper gripping, closed-loop control is implemented using the tactile sensors. Feedback for the closed-loop control is provided by force sensing resistors (FSRs), attached on the fingertips of the robotic hand. These sensors also enable handling of fragile objects. Index Terms - Imitation, dexterity, gripping, bio-mechatronic, artificial hand

I. INTRODUCTION

Human hand signifies a magnificent and challenging example for scientists and engineers trying to replicate its complex structure and functionality. Art ificial hand robots are intended to be anthropomorphically designed so that the natural feeling is exh ibited. Similarly robotic hands are designed having dexterity and functionality as close as the natural human hand produce. The Robonaut hand by NASA [1],GIFU hand [2], DLR robotic hands [3], The UB Hand Evolution[4] and others are notable research works and designs in the domain of artificial hand having human like functionality and dexterity. But the aforementioned work is highly expensive and complex which makes it unable to be applied in low-cost research projects and prosthesis. This cost issue motivates the development of a low-cost, effective anthropomorphic and dexterous hand that can be used for inexpensive prosthesis and basic research projects compared to the other costly systems. To add some research value to the present work, the system to be developed in this work is chosen to be under-actuated; a single finger with three joints is controlled by one actuator. The system’s kinematic model is approximated by using simple jo ints i.e. pin joints.

Approximation using pin joints is an effective means of modelling, as these are in fact the same as compared to proximal and distal joints of humans. But note that for metacarpal joints this pin joint approximat ion is not valid as it is more complex [5].Our anthropomorphic robotic (or simply ‘art ificial’) hand comprises of four under-actuated fingers and one under-actuated thumb. Each finger consists of three phalanges or links and these are connected to each other by using pin joints. Due to these three phalanges in each finger the robotic hand achieves a stable grip which also adapts in changing condition. [6, 7] Each finger in the robotic hand moves due to a cable mechanis m. Two cables run along the internal structure of the finger and are attached to the outermost phalange. [8] The other end of the first cable is attached to the actuator, and the other end of the second cable is attached to a spring which gives an extension to the finger, such that it comes back to its original position after movement due to an actuator. Also due to the cable mechanism used, the hand can adapt itself to the object shape. [9, 10] Our anthropomorphic robotic hand has five degrees of freedom (DOF) and two degrees of motion (DOM). Each finger is actuated by separate RC servo motors, placed in the fore arm region of the hand, as suggested in [11]. The dexterity of the artificial hand is exh ibited by imitating the natural motion of the human fingers. Imitation is achieved by two different methods; a camera based marker recognition system to identify the human hand gestures and acquired flexion data from sensors attached to the human fingers. In order to have proper gripping, closed-loop control is implemented using the tactile sensors. Feedback for the closed-loop control is prov ided by force sensing resistors (FSRs), attached on the fingertips of the robotic hand. These sensors also enable handling of fragile objects. The following sections discuss all of that in details.

II. MODELING AND DESIGN

Several techniques are available for dynamics analysis of the system. Computer simulation software, e.g. ADAMS®, can be used for the simulation of dynamics and kinematics of the artificial hand. Also the forward kinematics method is useful to estimate the position of the fingers in x-y plane, provided the length of the links and the angular displacement

10

2014 International Conference on Robotics and Emerging Allied Technologies in Engineering (iCREATE) Islamabad, Pakistan, April 22-24, 2014

978-1-4799-5132-1/14/$31.00 ©2014 IEEE

are known. This method is applied to calcu late the position of the fingers in a Cartesian x-y p lane, for any angular displacement . The change in is caused by the rotation of the motor attached to wire running through the finger. The location of a single finger in x-y plane is determined. Position of other fingers is found by similar approach. Finger scheme and its corresponding lengths and angles are shown in Fig #1.The lengths of the phalanges are given as , and

are the angles providing the angular displacement, where , and .

Displacement equations for each axis are given as follows:

The torque produced by the actuator can be related to the angular displacement by following equations:

,

These equations derived from the forward kinematics are used to implement the MATLAB® simulat ion for the artificial hand as shown in Fig # 2.

Fig # 2. Matlab Simulation of Robotic hand, based on derived kinematic equations

The torque required by the motor is calculated through empirical analysis. First the displacement in the spring for different loads is calculated through simple experiments. Then the values of spring stiffness constant K was found for applied loads shown in Fig # 3.

Fig # 3. Spring deflection

Deflection is direct ly proportional to mass in static equilibrium position. For static condition

The plot of the results is shown in Fig # 4. From this, mean value for K:

Δ = 2.7 cm (full bending of finger)

The circumference on which wire will be ro lled has a radius of 2 cm so the torque required is:

Y-axis

X-axis Fig # 1. Definition of angles and lengths

Y-axis X-axis

Z-axis

11

Fig # 4. Spring Constant K

Relation between the motor and the finger rotation is also found through empirical analysis. For this purpose, distance dbetween the joints at different rotation angle is calculated. At 0 rad with the y axis, d becomes maximum and at dbecomes minimum. Fig # 5 provides the illustration.

Fig # 5. dmax and dmin between joints

From the experiment, it can be found that

If denote the above distances for all three joints,

For Motor rotation,

From experimental analysis,

The 3D CAD model of the artificial hand is designed by using SolidWorks®. Analysis is done on the computer model and the modificat ions required to harmonize the design are made. Strength of the mechanical structure is also verified through

the dedicated feature of this software. Hand model is shown in Fig # 6.

III. DESCRIPTION

The under-actuated artificial hand is designed to perform basic movements and gestures of a human hand. Imitation of the human hand gestures is achieved through two different methods and the adaptive gripping was achieved through the closed loop control.

A. Flex Sensor Based Imitation In this method, imitation data is acquired from the flex

sensor shown in Fig # 7. The flex sensors measure the flex generated by the human fingers and the data acquired is then used to generate motion in the art ificial hand by actuating the motors. Five flex sensors are attached to the human hand through glove to produce human machine interaction.

Fig # 7. Flex Sensor

Empirical relation between output voltage and deflection is found to be a nonlinear relation as shown in Fig # 8.

Fig # 6. 3D CAD model of hand

12

Fig # 8. Nonlinear relation between deflection and O utput Voltage

Through a curve fitting function provided by Matlab®, the following affine equation can be derived:

where x is degree of deflection and y is output voltage.

B. Machine Vision Based Imitation In this technique a camera is used to detect the gestures

from human fingers, in Fig # 10 black coloring markers are shown which are used to identify the gestures. A Simulink® based algorithm is implemented to track the motion of the human hand, it is continuously applied on the images acquired from the video. First of all a single image of the real hand is acquired. Then the acquired RGB image is converted in to grayscale image. Then using the auto threshold block, the grayscale image is converted in to binary (black & white) image using a specific value of bandwidth. Now 5 d ifferent areas of markers can be seen on the video with some noise. Median filter is then used to reduce most of the noise and to get a clear picture. Now centroid of each area is calculated using the blob analysis block. Now after acquiring the centroids, this 2-D data (position of pixel in x-y p lane) is converted in to a single co lumn. These positions of centroids are now sent to the controller of the robotic hand step by step. A counter is used to make sure that each value is sent step by step. The movement of the thumb is measured by the change of its marker’s position in both x-axis and y-axis while for other fingers, the position is measured in y-axis only. These points are also sent to the draw marker block to mark the centroids in the video.

Fig # 9. Hand Glove and Markers

The resolution of the camera is 358×288. It means that total number o f p ixels in x-axis is 358 while 288 in y axis. The camera was placed at some distance (approximately 42 centimetres) from the human hand so that all the 5 five marker can be covered in this resolution with minimum distortion. Table # 1 shows different positions (pixel position) of all the 5 markers in x-y plane. This data is arranged in such a way that the 1st row shows a fully stretched while the last row showsthe fully deflected state of human hand.

C. FSR Based Feedback Control

For closed loop control, FSR (Force sensitive Resistors shown in Fig # 9) are attached on the finger tips. This module ensures safe and appropriate gripping response.

Fig # 10. Force Sensitive Resistor

IV. ANALYSIS

When mimicking is done through machine vision, it is evident from Table # 1 that, the 4 fingers are changing position in y-axis only whereas the thumb can have change of position in both x-axis and y-axis. Therefore interpret ing the position of the thumb of human hand in 2D, on the thumb of artificial hand is considerably complex than the interpretation of other four fingers. Also by applying the machine vision based mimicking technique only visual perception of the human hand movement is observed by the camera and since there is no physical interaction of sensor with the human hand, accurate measurement of human hand position is not possible. Therefore the accuracy and response of the mimicking is questionable in this case. Also the processing cost in this technique is much higher compared to the accuracy and response of the system. Whereas when imitation data is acquired through the flex sensors, the accuracy and response

1st

finger (Thumb)(x, y)

2nd

finger (Index finger)(x, y)

3rd

finger (Middle finger)(x, y)

4th

finger (ring finger)(x, y)

5th finger (little finger)(x, y)

1. 22, 188 137, 56 185, 29 260, 58

315, 96

2. 61, 205 138, 81 186, 60 257, 80

309, 111

3. 99, 219 138, 113

186, 99 253, 114

302, 137

4. 132, 231 139, 140

186, 143

250, 143

295, 160

5. 170, 239 139, 177

186, 174

247, 173

288, 176

Table # 1

13

of the system increases comparatively. Also the processing cost in this technique is considerably less and can be achieved through the microcontrollers only. It was observed that when both techniques were applied simultaneously, with one technique providing the data to be used as feedback for the whole system, the accuracy and response of the system is maximized. There are two types of grips generally conceived in artificial mechanical hands. These two types of grips are typically used to judge and test the capabilities and dexterity of a mechanical hand. These two types are Circu lar pinch grip and Pris matic power grip. It turns out that our low-cost artificial hand has the capability to produce both types of mentioned grips. Fig # 12 shows that our artificial hand can perform a circular pinch grip through a proper force being applied between the thumb and the tips of the fingers. Through this type of grip, our artificial hand can hold or grip thin objects i.e. p iece of paper, pencil etc. It is also possible to grip a large-diameter bottle.

Fig # 11. Circular pinch grip on a golf ball

Note that this circular p inch grip applies significantly lower pressure (Fig # 13) and it gets difficult to retain the grip on the object, especially when the weight of the object is significantly large. Our art ificial hand turns out to be capable of performing this kind of grip successfully.

Fig # 12. Circular pinch grip on a small motor

In prismat ic power grip, some force is applied between palm or base of the hand and the joints of the four fingers. This type of grip allows holding bottles and other circular objects with s maller d iameter. The prismatic power grip can apply higher pressure than the circular pinch grip. Fig # 14 shows that our artificial robot is also capable of performing this kind of grip. Note that the s mallest finger has a very trifling contribution to the applied pressure in both types of grips .

Fig # 13. Prismatic power grip on an object

It is demonstrated that our artificial hand is able to hold frag ile objects through tactile feedback shown in Fig # 11. As seen in the figures below, the mechanical hand successfully holds an egg without damaging it. By using FSR, the force required to hold an egg or other fragile objects can be calculated through experiments, and these experimental values can be used as threshold values to stop the application of force on the fragile objects.

V. CONCLUSION

In this paper, a simple and low cost mechanism of an anthropomorphic robotic hand capable of performing basic human hand motions with fundamental gripping functionality is proposed. Each finger of the robotic hand consists of three joints and is controlled by a single actuator, through a wire guided mechanis m. For a well achieved imitation of human hand gestures two methods are proposed including the imitation data from the flex sensors and the data received through the image processing based tracking of finger markers. By employing these two techniques simultaneously, the response and accuracy of the imitation can be increased. To achieve more realistic and adaptive gripping through the robotic hand, a closed loop control system based on the feedback from the force sensitive resistors was developed. The paper suggests an inexpensive and relatively simple techniques to accurately simulate human hand gestures, movements and gripping techniques for prosthesis and research experiments.

Fig # 14. Artificial hand holding an egg and mobile phone

14

REFERENCES [1] C.S. Lovchik and M.A. Diftler, “The Robonaut Hand: a Dexterous Robot

Hand for Space”, in Proc. IEEE Int. Conf. on Robotics and Automation ’99, ICRA 99, Michi- gan, May 1999.

[2] H. Kawasaki, H. Shimomura, Y. Shimizu, “Educational- industrial complex development of an anthropomorphic robot hand ’Gifu hand”’, Advanced Robotics, Vol. 15, No. 3, pp. 357-363, 2001.

[3] J. Butterfass, M. Grebenstein, H. Liu, G. Hirzinger, “DLR- Hand II: Next Generation of a Dextrous Robot Hand”, in Proc. IEEE Int. Conf. on Robotics and Automation ’01, ICRA 01, Seoul, Korea, May 21-26, 2001.

[4] C.Melchiorri and G.Vassura , “Mechanical and control features of the UB Hand Version II”, in Proc. IEEE/RSJ Int. Conf. on Intelligent Robots and Systems ’92, IROS 92, Raleigh, NC, 1992.

[5] I.A. Kapandji, The Physiology of the Joints: Upper Limb, vol. 1, 5th ed. New York: Elsevier, 1986.

[6] R. M. Murray and S. S. Sastry, A Mathematical Introduction to Robotic

Manipulation. Boca Raton, FL: CRC Press, 1993. [7] A. Bicchi, “Hand for dexterous manipulation and robust grasping: A dif-

ficult road toward simplicity,” IEEE Trans. Robot. Autom. vol. 16, no. 6, pp. 652–662, Dec. 2000.

[8] S. Hirose and Y. Umetani, “The development of soft gripper for the ver- satile robot hand,” Mech. Mach. Theory, vol. 13, pp. 351–359, 1978.

[9] R. Cabas and C. Balauger, “Design and development of a light weight em- bodied robotic hand activated with only one actuators,” in Proc. IEEE/RSJ Int. Conf. Intell. Robots Syst., pp. 741–746, 2005

[10] I.W.Park,J.Y.Kim,J.Lee, and J.H.Oh,“Mechanical design of humanoid robot platform KHR-3 (KAIST humanoid robot-3: HUBO),” in Proc. 5th IEEE-RAS Int. Conf. Humanoid Robots, pp. 321–3262005, 2005.

[11] M.C.Carrozza,C.Suppo,F.Sebastiani,B.Massa,F.Vecchi,R.Lazzarini, M. R. Cutkosky, and P. Dario, “The spring hand: Development of a self adaptive prosthesis for restoring natural grasping,” Auton. Robots, vol 16, pp. 125–141, 2004.

15