The Use of Intelligent Control in a Prosthetic Hand

P.J. Kyberd M. Evans


Introduction

There are few paths that can be used as part of the control of conventional artificial arms. They are generally visual, audio and occasionally, vibrational. In addition, conventional wisdom suggests that good control results from allotting one joint per input channel, [1]. The result of this is that the complex co-ordination of parallel joints is difficult to achieve by all but the most skilled or practiced operators. This limitation in the number of independent degrees of freedom the prosthesis can have means that the functional range will be reduced. An alternative strategy is to allow a computer based system to coordinate multiple axes and leave only the strategic control to the operator, [2,3]. The additional motions can then be employed to provide more functional grip forms in a single terminal device, and the computer controller can make the use and activation of the device more anthropomorphically appropriate.

Figure 1.

Intelligent Control

The control philosophy is generally referred to as the Southampton Adaptive Manipulation Scheme (SAMS), [4]. It is applied to electrically driven hand prostheses. Although it has only been used with a myoelectric input, the computer-based controller would allow for many other input methods. Conventional myoelectric controllers use two inputs on opposing muscles to drive the hand in opposite directions; Voluntary opening and closing. This ensures that some concentration must be devoted to the deliberate control of the flexion of the hand. For a SAMS hand the flexion of the hand is proportional to the level of tension from a single EMG channel. This is more similar to a voluntary opening involuntary closing, so that the user can open the device and then allow the hand to close around the object. Sensors in the fingers then detect the contact with the object and the controller stops the fingers automatically. The shape of the hand can then match the shape of the object, while the hand applies the lightest feasible touch to the object. At this point, if the operator wishes to move the target object within the hand, they may do so. The sensors will detect any change in force and advance or retard the fingers to maintain this force. The second muscle is used to switch the hand into a HOLD mode. Additional sensors on the hand detect the slippage and automatically increase the grip force without any intervention from the user. Alternatively, if the user wishes to override this reflex then further muscle tension is converted to a demand to increase the grip force proportionally. At anytime the tensioning of the original muscle will cause the hand to open. Thus the instructions are progressive, natural and consistent.

The intelligence of the device extends to the decision to open the hand. Conventional controllers generally cannot distinguish between the instruction to open the hand when empty and when it is holding an object, thus some form of compromise must be made. The threshold tension (and so EMG) level must be set in one of two ways. Either, it is too easy to open a held object (and so drop it), or too hard to open an empty hand (and so make the operation of the hand tiresome). A SAMS hand can make a distinction between the two actions.

Secondly, the shape the hand adopts can depend on the shape of the object. From the generic forms (precision and power) the hand can change its shape depending on the shape that the target object presents to the hand. For example, the hand starts in a precision grip, but if the object contacts with the palm of the object, the thumb flexes and the fingers close on the object. This provides a greater number of points of contact between the object and the hand. This reduces the required grip force and creates a more stable grip.

The concepts for the device are now quite mature, and while some of the above features have been realized on other hands no other device incorporates all of the above features in a single prosthesis. These ideas have been used with four, one, and most recently, two degrees of freedom, [5,6,7]. The ideas have in the past, been associated with a single device or class of prosthesis. This is, and was, a misapprehension. The computer controller allows it to be applied to many degrees of freedom and both cosmetic and non-anthropomorphic terminal devices. The current hand has been developed for the extended clinical usage of the control concepts in the field, and it is this version that is detailed herein.

The Oxford Hand

  1. The hand should operate without straps.
  2. Any new form of prosthesis must be light; power is a lesser concern.
  3. An externally powered hand must be capable of operating for over twelve hours, either on a single charge or by simple repowering.
  4. Any new device must be able to drive a car easily, preferable without any modification.
  5. New materials for gloves that are easily cleaned, hard wearing, cosmetic, and able to flex in more than one direction without deformation.

Two degrees of freedom were chosen for the prosthesis as a compromise between the advantages of multiple degrees of freedom and the weight and complexity penalty of such a device. Two degrees allow for the two major grip postures to be adopted. The device has to be a self-contained system with all the drives within the hand itself. It must be reversible, lightweight, and anthropomorphic in shape and action. Small to medium in size and capable of being opened when un-powered. The resulting design has undergone life trials. The hand's internal mechanism is symmetrically neutral so that the device is easily reversible. The drives actuate fingers and a thumb separately. The finger drive is braked so that the hand will not back drive during grasping, but the user can manually release the brake. The hand offers the synergetic prehension format with the major drive force coming from the fingers [9]. This also agrees with the findings of Wing [10,11] and ourselves that on the natural hand the thumb moves more slowly than the fingers during prehension.

The digits continually curl from straight at full extension to touching the palm at full flexion. This ensures a wider range of objects can be stably held in the palm, compared with fixed geometry devices. The distal phalanges of the fingers are also connected together via a force equalizing mechanism so that an irregular shaped object can still be grasped evenly in the hand.

The grip posture is more anthropomorphic in form and it allows the hand to have a greater area of contact with the object. It has been shown that for natural prehension the majority of grasps are below 10N. Most prostheses appear to have insufficient grip force. This is because the load is spread over a very small area of contact. If the grip is more conforming then the device will need a lower grip force for stable manipulation.

Controller

The electronics hardware is based around an Intel 80C196 microcontroller. The total electronics package is 2" by 2" by 1 V" and has input channels for three force, three slip and one touch sensors. In addition it has two EMG channels and measurement of the degree of freedom of the fingers. A serial link is available to a PC for calibration and diagnostics. The electronics draws a maximum of 50mA, with the low power consumption at 25mA. The target operational life of a battery is twelve hours on a single charge. The questionnaire of users showed that if users wear their arms they tend to wear them for up to 14 hours a day, [8]. It is this level of use the system is designed to deliver.

Training

The users are introduced to the concepts slowly and progressively. First, once the electrode positions are established, a link from the controller electronics is used to display the EMG levels input to the controller. This shows the two values as horizontal bars across a screen. They are asked to hit particular levels or use particular muscles in conjunction or separately. Once they have some confidence in their control they change to a simulation on the PC that displays a hand that operates similarly to the real prosthesis. Then the idea behind the control can be demonstrated. It has been a common experience that the conventional form of control has to be unlearned before the new control is achieved. It is, however, not a potentially difficult task. This is clear in general, as it is not uncommon to make a bilateral fitting with a myoelectric hand on one side and a body-powered device on the other. So the juxtaposition of voluntary closing and involuntary closing, is an idea that many users can accommodate. The hand is operated on the bench before it is fitted to the arm. Further training of the device using standard tasks and objects, follows a similar pattern to conventional devices. In addition, foam objects are used, and they are arranged so that the grip force imparted on the object is made clear to an observer. This is achieved by gluing strips of foam together in the middle so that when they are squeezed they splay out and the force level is made visible.

Acknowledgements:

This project is funded by the Leverhulme Trust and the Cloth-worker's Foundation.

Oxford Orthopaedic Engineering Centre

Nuffield Orthopaedic Centre Headington, Oxford UK OX3 7LD

References:
1. Radonjic, D., and Long C, Why Myo-electric control is so difficult Proceeedings of the symposium on external control of human extremities, Dubrovnik, August 1969,pp 59-67. (1969)
2. Baits J.C., Todd R.W. and Nightingale J.M. The feasibility of an adaptive control scheme for artificial prehension In : Proceedings of a symposium at the I.Mech.E. 183-3J:54-59. 1968
3. Kyberd, P.J , and Chappell, P.H., The Southampton Hand, An Intelligent Myoelectric Prosthesis Journal of Rehabilitation Research and Development, 31, 4, pp 326-334, 1994.
4. Chappell, P.H., and Kyberd, P. J., The prehensile control of a hand prosthesis by a microcontroller J. Biomed. Eng.m 14, September, 1991.
5. Chappell, P.H., Nightingale, J.M., Kyberd, P.J., and Barkhor-dar, M., Control of a single degree of articial hand J. Biomed. Eng.,9,July 1987,273-277.
6. Kyberd, P.J., Carnegie, F., Mustapha, N., and Chappell, P.H., Clinical experience with a hierarchically controlled, myoelectric hand prosthesis with vibro-tactile feedback Pros. And Or-tho Int. 1993, 17,56-64.
7. Kyberd, P.J., Holland, O., Chappell, P.H., Smith, S., Tregidgo, R., Bagwell, P.,and Snaith, M., MARCUS: A Two degree of freedom Prosthesis with Hierarchical Grip Control. Ieee Trans. Rehab. Eng. March 1994. Volume 3 no 1 p 65-69.
8. Kyberd, P. J., Survey of the population of upper limb users in the Oxfordshire region. International Society of Pros. And Or-tho UK annual meeting, Swansea 31st March-2nd April 1993.
9. Prout W. The New York electric elbow, the New York prehension actuator and the NU-VA synergetic prehensor Chapter 20, The comprehensive management of the upper-limb amputee. D.J. Atkins and R.H. Meier, editors, Springer-Verlag. (1988).
10. Wing, A.M., and Fraser, C, The contribution of the thumb to reaching movements Quarterly J. of Exp. Psychology.35A, 297-300.1987.
11. Wing, A.M., and Fraser, C, Artificial hand use in grasp and release phase of reaching International symposium of teleopera-tion and control. Bristol, July 1988.