About an antropomorphic robotic hand.

Stanciu, Loredana ; Stanciu, Antonius

1. INTRODUCTION

From all the components of the human body, the hand is the most interesting, due to its complex structure and its vital role in daily activities. The hand is a very good example of how to implement an impressive body part, able to fulfill elaborate and useful tasks, using a combination of mechanisms, senses, sensors, and control functions. It is not just a simple tool, but an ideal data acquisition instrument from the surrounding environment.

There are quite few solutions for robotic hands, based on mechanical, electrical, electromechanical, pneumatic, or hydraulic implementations. The models proposed so far--LMS Mechanical Hand (Chaigneau et al., 2008), Robonaut Hand (***, 2007), DLR Hand (Butterfass et al., 2001), Barret Hand (Yang et al., 2004), Shadows Dexterous Hand (***, 2008)--represent acceptable solutions, but most of them are too complex and difficult to control. Comparing the hand's performances with the one of the existing artificial models, the results can be very disappointing. Still, this challenge determined the researchers to continuously try to copy it.

The interaction between the hands and the objects is made with a dexterity which implies a strict control not only of fingers' final location, but also of joints' angles, forces, and (maybe the most important) of fingers' rigidity. To copy all these human characteristics into a robotic hand, the most direct (and, probably, optimum) approach is the duplication of the anatomical musculoskeletal structure of the human hand (Matsuoka et al., 2006).

[FIGURE 1 OMITTED]

In order to have a functional robotic hand, it is very important to design it primarily for grabbing. After that, one can improve the design for manipulation which assumes high dexterity, advanced sensors, complex control technologies (Carrozza et al., 2004). Also, placing the actuators outside the hand will determine a lighter and simpler structure for the fingers. In this case, actuators with a higher power can be used (pneumatics or hydraulics), so the resulting grabbing force will be higher. On the other hand, the structure is not overloaded with the actuators' weight (placed on the forearm, for example).

Based on these observations, this paper presents a working prototype of an anthropomorphic robotic hand. The motors are placed outside of the hand and a hydraulic actuation can be used. In this early stage it is not feasible to implement a very complex model, so the prototype is capable to perform only the prehension function. This solution led to a cheap and simple model which can be successfully used to grab objects.

2. METHODS

Prior to designing the prosthesis, one has to study the natural model to observe its functionality and its structure. The human hand is a highly articulated system, composed of five fingers (four with the same structure, and the fifth, the thumb, with a slightly different one) having 23 DoFs. The motion of this system is submitted to some constraints (Lin et al., 2000), so the hand cannot make arbitrary gestures.

The kinematical structure of the human hand can be seen in Fig. 1. The general coordinate system to express the motion of the hand is placed at the wrist level. The kinematic model is slightly different than the natural one having only 22 DoFs. The explanation is that the thumb has lost one DoF because of the way the model was designed in order to make the resulting equations easier to be determined and easier to work with. The model is still able to grab things because the thumb is placed on the palm in such a way to not disturb this important function of the hand.

The proposed model implements, for now, only the prehension function. The reasoning behind this design decision is to considerably simplify the model, so it will offer an acceptable economical efficiency. To obtain the mechanical linearization of the ratio between the generated force and the grabbing force and to solve the difficult problem of the constrictive actuator (the so called artificial "muscle"), needed to move the artificial hand's phalanges, a simple, yet effective, mechanism was created (Stanciu et al., 2006).

The robotic hand has a very similar structure with the natural model (Fig. 2). There are five fingers: four of them have three phalanges each and the fifth, the thumb, has only two phalanges. Having in mind to implement only the prehension function, a degree of freedom (DoF) from the metacarpophalangeal joint, the one concerning the adduction or abduction motions, was eliminated. To ameliorate this loss, the four central fingers were linked to the palm to assure the grabbing of objects having various shapes. With this constraint, the model has 14 DoFs, plus the 3 DoFs of the wrist.

[FIGURE 2 OMITTED]

Studying the way the natural hand grabs the objects, it can be seen that the phalanges wrap around the objects trying to follow their shape. Also, when grabbing, the human perception does not pay attention to the exact position of each phalanges, but focuses on the tactile sensation instead. This way, dangerous objects are not grabbed and no object is squeezed beyond its breaking point.

Hydraulic actuators were used to move the artificial hand. This allows the placement of the motors on the forearm and the elimination of the tactile sensors from the finger level. To control the grabbing force, pressure sensors, placed also on the forearm, can be used (considering that the pressure of the fluid in the hydraulic circuit is statically the same in any point).

When there is a need to grab an object, the control circuitry will issue the command to the pumps for increasing pressure. A pressure limit should be set for every phalange, according to the nature of the object to be grabbed. Every phalange will begin to close around the object, without the necessity of knowing its current position. The movement will stop automatically when all the pressure limits were reached. This will be the moment when the object is considered grasped and can be moved.

Stepper motors were used to drive the artificial hand. There are several reasons supporting this decision (Ungureanu et al., 2006), the most important being that they generate a high torque at low rotational speeds, which allows precision motion of the phalange. To solve the delicate problem of transforming the rotational motion of the motor's shaft into a translational motion, UBL23 unipolar stepper motors were actually used. A data acquisition board from National Instruments (NI-PCI-6509) was used to take over the signals from a computer and send them to motors' drivers.

The actions of the artificial hand are governed by software running on a computer. This software sends the driving signals to the motors and receives signals from the sensors through the data acquisition board. The technological limitations, encountered when implementing the artificial hand's structure and the driving part, led to a simplified version of the program (Ungureanu et al., 2007). The resulting structure does not use sensors (the pressure in the system has pretty low values and the resulted force will not be able to break an usual object).

3. CONCLUSION AND FUTURE WORK

When copying a natural model, it is always important to carefully study its structure and characteristics. In the case of human body, this is very difficult because of its complexity, so it will result in a simplified artificial model. When talking about the human hand, not only to reproduce the kinematical structure is a challenge, but also to study its dynamic behavior, absolutely necessary because its normal physiological motions require dynamics. Another problem is to choose the appropriate actuators able to assure the laws of motion described by the kinematical equations (when the dynamic model has an important role) and to manufacture the phalanges and the joints as anatomically similar as possible, and to made them out of light but strong materials.

What we obtained was a working prototype for a robotic anthropomorphic hand, able to perform the prehension function. Although the model can grab various objects in size and shape, it still needs a lot of improvements. The main drawback is the lack of pressure sensors in the system, which does not assure a good control of the grab. This is because the pressure in the hydraulic system is low (but enough to grab objects lighter than 1 Kg without damaging them) due to some implementation problems.

After solving the mentioned difficulties, some grabbing patterns should be established (because, when grabbing, the hand adjusts to a certain shape, but the exercised pressure should not exceed what the object can withstand without breaking or deforming).

4. REFERENCES

Butterfass, J.; Grebenstein, M.; Liu, H. & Hirzinger, G. (2001). DLR-Hand II: Next Generation of a Dextrous Robot Hand, Available from: http://www.dlr.de/rm/Portaldata/52/ Resources/dokumente/hand/icra2001next.pdf Accessed: 200910-14

Carrozza, M.C; Suppo, C; Sebastiani, F; Massa, B; Vecchi, F; Lazzarini, R.; Cutkosky, M.R. & Dario, P. (2004). The SPRING Hand: Development of a Self-Adaptive Prosthesis for Restoring Natural Grasping, Autonomous Robots, Vol. 16, No. 2, (March, 2004), pp 125-141, ISSN: 0929-5593

Chaigneau, D.; Arsicault, M.; Gazeau, J.P. & Zeghloul, S. (2008). Lms robotic hand grasp and manipulation planning (an isomorphic exoskeleton approach, Robotica, Vol. 26, No. 2, (March, 2008), pp 177-188, ISSN: 0263-5747

Lin, J.; Wu, Z. & Huang, T.S. (2000). Modeling the Constraints of Human Hand Motion. Workshop on Human Motion, pp 121-126, ISBN: 0-7695-0939-8, Los Alamitos, USA, 2000

Matsuoka, Y.; Afshar, P. & Oh, M,, (2006). On the Design of Robotic Hands for Brain-machine Interface, Neurosurgery Focus, Vol. 20, No. 5, (May, 2006), pp 1-9, ISSN: 10920684

Stanciu, A.; Dragulescu, D. & Ungureanu (Stanciu), L. (2006). A Hydraulic Solution for Implementing Human's Hand Prehension Function, SIITME 2006, International Symposium for Design and Technology of Electronic Packaging, pp 207-210, ISBN: 978-973-8961-23-4, Iasi, Romania, September 21-24, 2006

Ungureanu (Stanciu), L.; Stanciu, A. & Menyhardt, K. (2006). A Hydraulic Solution for Actuating a Human Hand Prosthesis. WSEAS Transactions on Systems, Vol. 6, No. 1, (January, 2007), pp 40-46, ISSN: 1109-2777

Ungureanu (Stanciu), L.; Robu, N. & Manoila, V. (2007). Driving Software for an Artificial Human Hand Hydraulically Actuated, The Scientific Buletin of Politehnica University of Timisoara, Vol. 52 (66), No. 3, (September, 2007), pp 131-134, ISSN 1224-600X

Yang, J.; Pitarch, E.P. & Abdel-Malek, K. (2004). A Multi-fingered Hand Prosthesis. Mechanism and Machine Theory, vol. 39, no. 6, (June, 2004), pp 555-581, ISSN: 0094114X

*** (2007) http://robonaut.jsc.nasa.gov/media/videos/videos .asp#hand_motion, Accessed on:2009-10-14

*** (2008) http://www.shadowrobot.com/downloads/ shadow_dextrous_hand_technical_specification_C5.pdf, Accessed on:2009-10-14