DESIGN, ACTUATION, AND CONTROL OF A COMPLEX HAND MECHANISM. by Jason Dean Potratz

Transcription

1 DESIGN, ACTUATION, AND CONTROL OF A COMPLEX HAND MECHANISM by Jason Dean Potratz A thesis submitted in partial fulfillment of the requirements for the Master of Science degree in Mechanical Engineering in the Graduate College of The University of Iowa July 2005 Thesis Supervisor: Professor Karim Abdel-Malek

2 Copyright by JASON DEAN POTRATZ 2005 All Rights Reserved

3 Graduate College The University of Iowa Iowa City, Iowa CERTIFICATE OF APPROVAL MASTER S THESIS This is to certify that the Master s thesis of Jason Dean Potratz has been approved by the Examining Committee for the thesis requirement for the Master of Science degree in Mechanical Engineering at the July 2005 graduation. Thesis Committee: Karim Abdel-Malek, Thesis Supervisor Jasbir Arora Shaoping Xiao

4 Sorry doesn t put thumbs on the hand, Marge. - Homer J. Simpson ii

5 ACKNOWLEDGMENTS I would like to that Professor Karim Abdel-Malek for supervising this research and serving on my thesis committee. I would like to thank Professors Jasbir Arora and Shaoping Xiao for also serving on my thesis committee. I would like to thank Dr. Timothy Marler for thesis editing and expert writing advice. I would like to thank Dr. Jingzhou Yang for collaborating on this work. Finally I would like to thank my friends and family for their support. iii

6 TABLE OF CONTENTS LIST OF TABLES LIST OF FIGURES v vi CHAPTER I. INTRODUCTION Literature review Motivation and objectives Overview of thesis 9 II. MECHANICAL DESIGN Design of the spring deflection joint mechanism Design of the actuation system Finger deflection Aesthetics 29 III. ELECTRICAL DESIGN Performance requirements Stepper motors versus servomotors Stepper motor theory and operation Stepper motor, drive, and power supply setup 45 IV. MOTION CONTROL Motion control setup Programming 55 V. APPLICATIONS AND RESULTS Applications Results 65 VI. CONCLUSION Contributions Shortcomings and future work 73 REFERENCES 77 iv

7 LIST OF TABLES Table 2.1 Spring properties 18 Table 2.2 Joint range of motion 27 Table 3.1 Force and torque requirements for each joint 33 v

8 LIST OF FIGURES Figure 2.1 Finger segment with cable and conduit 11 Figure 2.2 Degrees of freedom of fingers 12 Figure 2.3 Completed finger assembly 13 Figure 2.4 Cable routing 14 Figure 2.5 Hand anatomy 15 Figure 2.6 Entire hand assembly 16 Figure 2.7 Actuation system 19 Figure 2.8 Pulley assembly with shaft extension, coupler, pulleys, bearing, and nut 22 Figure 2.9 Fixture assembly with motors, motor drives, and pulley assemblies 24 Figure 2.10 Trajectory and range of motion experimental setup 25 Figure 2.11 Fingertip trajectory and orientation 28 Figure 2.12 Artificial and human hand, palm 30 Figure 2.13 Artificial and human hand, back of hand 30 Figure 3.1 Unipolar stepper motor 38 Figure 3.2 Bipolar stepper motor 39 Figure 3.3 Torque versus position for a single winding and two winding motors 41 Figure 3.4 Torque versus position for two windings using micro stepping 41 Figure 3.5 Current through winding versus time 42 Figure 3.6 Torque versus speed 43 Figure 3.7 Current versus time with and without current limiting 45 Figure 3.8 Manufacturer s torque rating (oz-in) of motor versus speed (revolutions/second) 46 vi

9 Figure 4.1 Actuation and control setup 50 Figure 4.2 Wiring diagram 52 Figure 4.3 State activation chart 56 Figure 5.1 Cylindrical grasp without glove 66 Figure 5.2 Spherical grasp with glove, one 66 Figure 5.3 Spherical grasp with glove, two 67 Figure 5.4 Pinch grasp with glove, one 67 Figure 5.5 Pinch grasp with glove, two 68 vii

10 1 CHAPTER I INTRODUCTION An artificial human-like hand based on a new design would fulfill different needs. There are numerous robotics applications that would benefit from an anthropometrically shaped devise, a gripper, capable of grasping and manipulating objects. Other grippers, often as simple as a one degree of freedom (DOF) pinching device, are used successfully in many applications, but a gripper that accurately reproduces human motion and grasping postures is novel and would be beneficial. For instance, in cases where the environment prevented or inhibited a human from completing a task, a robotic hand could be used as part of a system to complete the task. If humanoid robots were ever to become a reality, then a human-like hand would become a necessity. These robots would interact in the same environment as humans and therefore manipulate the same objects that would be designed with the intention of being manipulable by human hands. Another central need that would be fulfilled by a new artificial human-like hand would be advanced prosthetics. The ideal replacement for a missing human hand that accurately mimics a human hand in appearance, dexterity, tactile feedback to the user, and simplicity of use, remains far from a reality. Currently available myoelectric robotic prosthetic hands, which typically only have a few DOFs fall far short of this ideal prosthetic hand in terms of appearance, functionality, and easy of use. Surveys conducted on user satisfaction of myoelectric hands show that percent of users do not use their hands regularly. The reasons for this lack of use are: low functionality, unsatisfactory cosmetics partly caused by unnatural motion, and low controllability.

11 2 From a mechanics perspective, even these relatively simple hands are not natural to control; users have to exert an enormous amount of concentration for even simple tasks (Massa et al, 2002). In this work we developed a five-fingered hand with 15 DOFs that mimics a human hand in terms of appearance, motion, and grasping ability. One key feature that makes this hand more advanced than any other multi-fingered hands is the unique mechanism used for the joints. This newly developed mechanism is based on the loading of a compression spring in both the axial and transverse directions and offers several advantages over a hand mechanism based on rigid links and revolute joints. Since the structure of each finger is primarily made up of compression springs, the majority of the volume of each finger is occupied by empty space. Thus, the potential for constructing a very light hand exists if the proper materials are used along with an efficient design. A lightweight design is a critical aspect of a comfortable, wearable prosthetic devise and offers obvious advantages for robotics applications, such as smaller actuators and less power consumption. The unique mechanism also allows the hand to be inherently compliant, which enables the fingers to naturally conform to the shape of the object the hand is grasping, making for a more secure grip without the necessity of deliberately adjusting the defection of each joint to achieve the same shape. The joint mechanisms are actuated by a series of cables and conduits that allow the motors that actuate the joints to be located off the hand, also increasing the potential for a very light weight hand.

12 3 1.1 Literature review Some of the first mechanical hands were part of body-powered artificial limbs with a mechanical hook or claw type end effecter. Powered mechanical hands have seen some significant advances since these hands were considered state of the art. These advances have occurred primarily in prosthetics applications. An ideal solution for replacing a human hand is still far from being realized. Possibly one of the reasons this is true is that the two primary design criteria for prosthetic hands are often conflicting ones. Because the hand is part of the body, it is unique to each person. Likewise, the ideal replacement for the human hand would also be unique to each user. On the other hand, any prototype device has to be a solution that is well accepted by enough users to warrant developing it into a marketable product (Kyberd et al, 2003). Although there are many advanced hands under development, none of them incorporates all of the advancements that research has produced. Many of them implement advanced control schemes and sensors to achieve a significant level of automation, but many of the most advanced hands still do not have five powered fingers (see Butterfass et al, 1998, Tura et al, 1998, Carrozza et al, 2003, Zecca et al, 2003, Kyberd and Chappell, 1994), and Kyberd et al, 2001). None of them has as many fully actuated DOFs as the human hand, an attribute which of course, an ideal replacement for the human hand would have. These deficiencies are probably due to compromises made to save weight and to reduce the complexity of controlling a greater number of DOFs. Many commercially available hands, such as the Otto Bock SensorHand TM, only have three digits, usually the thumb, index, and middle fingers, and have two or three

13 4 DOFs, which are coupled together, and actuated by one motor. Hands such as these are only capable of being used for a precision or power grasp and have only limited functional use (Yang et al, 2005). One of the projects that has been under development the longest is the Southampton Hand at the University of Manchester Institute of Science and Technology. The Southampton Hand generically refers to the concepts that have been implemented in a series of hands, but the original was built in 1969 (Kyberd and Chappell, 1994). The earliest designs were quite large and relied on very simple electronics. Great potential for advancement in robotic hands came with the advent of the microprocessor. This was the key to solving many of the problems that prevented early hands from being very practical: size, weight, and power consumption. In the current state the Southampton Hand has two independent degrees of freedom, flexion of the first two fingers and flexion of the thumb. Combined force and slip sensors in the tips of the fingers along with a force sensor in the palm are used to aid the user in grasping and holding an object with minimal user input. The idea of minimal user input is one of the key concepts being developed as part of the Southampton Hand project, or more specifically that of hierarchical control. The user only has to provide higher-level commands like open, hold, or squeeze, and the movements of individual finger joints are coordinated by microprocessors using feedback from the sensors. This reduces the mental load on the user, which is a common cause of rejection of a device by the user (Kyberd, 2000). Another research project that has been under development since the mid 1980 s and has been a test bed for many other research projects to develop control strategies is the Utah/MIT Dexterous Hand. This hand has three fingers and a thumb with four

14 5 degrees of freedom each for a total of sixteen degrees of freedom. The joints are each powered by opposing extensor and flexor tendons, which are actuated by pneumatic actuators that are located in a remote actuator pack. There are Hall effect sensors to provide angular position feedback and the tendons pass over force sensors in the wrist to provide force feedback. The control system for the Utah/MIT hand attempts to control forces between the fingers and an object by recursive refinement of trajectories and by varying stiffnesses of the joints. Ultimately, controlled compliance of the fingers combined with a regulated position error produces appropriate forces to maintain the grip and to move the object within the grasp (Speeter, 1990). This enables the hand to grasp an object and manipulate it within its grasp. Although this hand is capable of complex manipulation and has served for a very effective test bed for developing other components and control schemes the bulky apparatus required to control it, including the pneumatic actuators, preclude it from being used as a prosthetic and give it a rather limited potential for use in practical robotic applications. A group in Italy is responsible for the development of a three fingered, twodegree of freedom, underactuated prosthetic hand. The key feature of this hand is the capability of an adaptive grasp due to the underactuation. Underactuation refers to a mechanism that has fewer actuators than it has degrees of freedom. The other degrees of freedom are then actuated by elastic elements, usually springs, and mechanical stops. The basic mechanism to control an underactuated system is a differential gear system. The mechanism employed in this hand is based on a tendon-like transmission system. The tendons are actuated by electric motors. The tendons then wrap around a series of three pulleys, one pulley being located at each of the three finger joints. The tension in

15 6 the tendon is then transformed into a moment that is proportional to the radius of each pulley and actuates the corresponding joint in the finger. The three joints in each finger are also fitted with springs to actuate the remaining degrees of freedom. The kinematic relationship between each link in the fingers can be modified by changing either pulley size or spring stiffness. This arrangement allows for an adaptive grasp that adjusts the shape of each finger to conform to the shape of the object being grasped (Massa, et al, 2002). This concept of adaptive grasping has shown much potential for grasping tasks in prosthetic applications but is not capable of manipulation tasks and is therefore unsuitable for most robotics applications. It will also have to be abandoned in order to someday achieve the perfect solution for a human hand replacement, which of course, would not be capable of just performing grasping tasks but also performing manipulation tasks. This would, of course, only be possible provided that a user interface advanced enough to be capable of interpreting the users intentions with minimal conscious effort is developed. The TUAT/Karlsruhe Humanoid Hand, which is being developed with both prosthetics and robotics applications in mind, is a five-fingered hand with twenty-one degrees of freedom that are driven by one actuator. This hand has a unique link mechanism to couple all of the joints together to allow for an adaptive grasp. Basically, all of the fingers begin to move as the actuator begins to move and the grasp begins to close. As one finger comes into contact with the grasped object that finger stops moving and the fingers not in contact with the object yet, continue to displace until they too come into contact with the object. In this way the individual fingers conform to the shape of the object, contact force is balanced between the fingers, and a truly adaptive grasp is

16 7 achieved. The fingers even tend to automatically readjust if the object inadvertently shifts within the grasp. Because of this, the researchers plan not to rely on any sort of sensors for feedback (Fukaya, et al, 2000). Again, using a hand with an adaptive grasp offers an advantage in the fact that it can achieve a stable grasp without the use of a complex control system, but also has a major disadvantage in the fact that this kind of hand can only ever be used for grasping and is not capable of performing any kind of manipulation tasks. 1.2 Motivation and objectives There remains a need to have a much more human-like artificial hand for many applications. Patients who lack part of an upper limb desire to not only restore the functionality of that limb but just as importantly any device used should be life-like and aesthetically pleasing. Currently available prosthetic devices fall short in many ways. They are hard to control and do not result in a natural motion. Due in part to their limited number of degrees of freedom, currently available prosthetic hands have limited practicality. They cannot perform a variety of tasks that would normally be completed using a human hand. Lastly, they are not aesthetically pleasing. Ideally the device would perfectly mimic the human hand in strength, dexterity, appearance, and easy of use. It should also be so natural to operate that the user would not have to put any more conscious effort into it than a real hand. This idealization is a far from being realized and there is much room for improvement in models currently available and under development. Even when considering projects under development, compromises have

17 8 been made in every prototype. Many still do not have five digits. Some are relatively simple to operate due to their adaptive grasp designs, but these can only be used in grasping type applications and cannot be used in any sort of application that requires complex manipulation. Currently available prosthetic devices lack degrees of freedom and powered digits. They cannot perform the same manipulation tasks and grasping postures as a human hand. They are cumbersome and impractical because they require a substantial amount of concentration from the user to operate and can only be used for a limited number of basic tasks. There is also much room to improve the aesthetics of these devises, which is an important part of overall comfort for the user. There is also a need to have a very human like manipulation ability in many robotics applications. For instance, in the remote operation of a robotic hand in situations where the environment either inhibits a human to do the same task unaided or prevents it all together. One specific example of this is the ADAH Project that involves developing a robotic hand to assist astronauts during extra vehicular activities in space who are normally hindered by the need to wear pressurized gloves (Carrozza, et al, 2002). Another instance would be in the eventual development of a humanoid robot which would likely be required to interact in the same environments as humans do on a regular basis, and therefore manipulate the same objects that humans do. The most logical end effecter for these kinds of tasks would, of course, be a hand that mimics the human hand. The objectives to be demonstrated as fulfilled by this research are: 1. Design and produce a five-fingered, fifteen degree-of-freedom prototype hand that implements a novel joint mechanism based on the deflection of a

18 9 compression spring. Mimic a human hand in size, appearance, and motion. Design the hand to accommodate a cosmetic glove. 2. Create a bench-top actuation and motion control system to explore, develop, and demonstrate the capabilities of the hand. 3. Develop software for a user interface to coordinate hand movement. 1.3 Overview of thesis Chapter 2 covers the mechanical design of the hand, including the following: design of the spring deflection joint mechanism, design of the actuation system, results of experimentation, and overall esthetics. Chapter 3 discusses the electrical design, including: performance requirements, how the stepper motors used in this application fulfill those requirements and general theory of operation of stepper motors, and will discuss the other electronic components used. Chapter 4 discusses computer control of the hand including the motion controller card used and the program developed for control of the hand and simulation of a user interface scheme under development that uses myoelectric signals to control a prosthetic device. Chapter 5 discusses use of the hand. The first section in this chapter covers potential applications of the hand, such as an improvement over current designs in prosthetic hands or the possibility of its use in robotics applications. The second section will discuss examples of what the hand is capable of. Chapter 6 provides a summary of contributions and a discussion of potential future work.

19 10 CHAPTER II MECHANICAL DESIGN Chapter two will cover aspects of the mechanical design of the hand and the system that actuates it. More specifically, it covers the novel flexing elements used to make up the joints of the hand, the actuation system, experimentation on the hand to determine the kinematic relationship between the orientation of the motor rotor and the position of the end effecter, and how the components were designed to be esthetically pleasing. These flexible joint elements are the basis for hand motion and at the same time make up the majority of the structure of the fingers. The purpose of the actuation system is to move the joints of the hand in response to computer-generated commands. It includes five stepper motors, fifteen flexible cable and conduit sets to transmit motion from the motors to the joints of the hand, and five pulley assemblies. 2.1 Design of the spring deflection joint mechanism The key feature of this hand mechanism that makes it unique is its joint mechanism used for the finger joints. Unlike all other mechanical hands, which employ solid links with revolute joints or another conventional mechanism, this one makes use of a flexible element made up primarily of a compression spring. Most of the structure for each finger segment is made up by this spring, Figure 2.1. It is also this flexible element that allows for the motion for the flexion and extension DOF of the finger, Figure 2.2. Each finger is made up of a series of three springs connected in series that are held in

20 11 place by short aluminum finger segment links. Each segment, in the current design, has one DOF and is actuated by a single cable. Although each joint currently has only one DOF to replicate the extension and flexion of the human finger, more cables could be added at appropriate orientations around certain finger segments to add additional DOF s; the DOF s would replicate the abduction and adduction motion of the human finger. Figure 2.1 Finger segment with cable and conduit Figure 2.1 depicts a CAD model of a typical joint segment that would be used to make up a finger. This segment includes a spring and the cable and conduit used to actuate the segment. Two aluminum finger segment links at either end are used to connect the spring to the springs in the preceding and following segments and to retain the ends of the cable and conduit. The finger segment links have a thread-like structure

21 12 at each end that screw into the first coil of the compression spring, holding the spring securely. Figure 2.2 Degrees of freedom of fingers Source: Peña Pitarch, E., Yang, J., Abdel-Malek, K., (2005) SANTOS TM Hand: A 25 Degree-of-Freedom Model Proceedings of the Society of Automotive Engineers Digital Human Modeling for Design and Engineering Symposium, June 14-16, Iowa City, Iowa The joint segment also includes a rectangular rubber block, or stiffener, oriented parallel to the axial direction of the compression spring, located at on the edge of the segment opposite of the cable. The finger segment structure offers very little resistance to lateral deflection but offers a much higher resistance to axial compression, which ensures that segment will deflect laterally in the direction of the palm (flexion of the finger) with very little resistance. The finger segment will require much more applied force to deflect in the direction of the back of the hand (extension of the finger). In other

22 13 words, it makes the finger segment stiffer when bent in one direction as apposed to the other. Figure 2.3 shows one entire finger unit with three finger segments and three cable-and-conduit sets to actuate all the segments. The tip of the thumb is located towards the right of the Figure 2.3. At the left of the Figure 2.3 is the thumb base, which slips into the aluminum hand body fixture. The other four fingers are constructed in a similar fashion. Figure 2.3 Completed finger assembly The cables and conduits run from the finger segment that they actuate, through the empty space in the center of the preceding elements, continue through the hollow finger bases, and exit though the bottom of the hand body fixture, Figure 2.4 and Figure 2.5. This arrangement allows for a single segment to be actuated without affecting the other two segments located in that finger. Figure 2.4 shows how a set of cables and conduits would be routed through a series of finger segments.

23 14 Figure 2.4 Cable routing The hand body is also constructed of aluminum and designed to resemble the shape of a hand. The fingers are arranged in a configuration that resembles human anthropometry and is suitable for grasping. Fifteen finger segments similar to one shown in Figure 2.1 were developed to comprise an entire five-fingered hand with three segments in each of the four fingers and the thumb to make a total of 15 DOF s. The thumb is capable of bending in extension and flexion at the carpometacarpal (CMC), the metacarpophangeal (MCP), and the interphangeal (IP) joints. Each finger is capable of bending in flexion and extension at

24 15 the metacarpophalangeal (MCP), the proximal interphalangeal (PIP), and the distal interphalangeal (DIP) joints, Figure 2.5. Figure 2.5 Hand anatomy Source: Peña Pitarch, E., Yang, J., Abdel-Malek, K., (2005) SANTOS TM Hand: A 25 Degree-of-Freedom Model Proceedings of the Society of Automotive Engineers Digital Human Modeling for Design and Engineering Symposium, June 14-16, Iowa City, Iowa Each finger can be added to and removed from the rest of the hand as one unit, which allows the entire hand to be assembled inside a cosmetic glove, which will be discussed more in section 2.4. Figure 2.6 shows the complete hand with three DOF s for each of the five fingers, giving it 15 DOF s.

25 16 Figure 2.6 Entire hand assembly One of the most important objectives when considering the design of the entire hand as a system is the ability to grip an object. The inclusion of a rubber stiffener in the finger segment is the key to making it operate in a way that makes gripping possible. Although the performance of the flexing element would be greatly reduced, it would be possible to use it without the stiffener. Without the stiffener, the stiffness of the compression spring itself represents an important tradeoff in the design of the finger segment. The maximum normal force that can be generated between a finger and an object grasped between the fingers and the palm or between the fingers and the thumb is proportional to the stiffness of the springs in that finger. When the hand closes to grasp the object, the cables shorten and pull the upper link towards the lower link, causing the finger segments to bend and curl towards the palm. After a finger first comes into contact with an object and more tension is placed on the cable, normal force is developed

26 17 between the finger and the surface of the object. Given that the grasped object exerts a force on the finger, there is a corresponding equal and opposite force for the force that the finger exerts on the grasped object. This force will cause the flexing elements to compress on the backside, opposite of the cable, causing the flexing element to simultaneously straighten and shorten. This effect produces undesirable and unrealistic motion and grasping postures and does not allow for sufficient force to be placed on the grasped object. Increasing the stiffness of the springs will increase the maximum amount of normal force that the finger is able to produce and also the maximum grasping force of the entire hand. Greater tension for actuating the hand would be required and therefore necessitate the use of large and more powerful motors for actuation. Using a rubber stiffener to increase the stiffness of the compression spring in only one direction is an innovative solution that simultaneously takes advantage of using both a stiffer spring which can create greater grasping force and a softer spring which is easier to actuate. This is because the material in between each link of the spring makes it much harder for that side of the spring to compress. This design still allows the fingers to react naturally to applied external forces. More than one set of springs was tested on the hand. The first set proved to be too stiff and would require using larger motors. A second set of springs that were softer was tested. This set proved to be much more practical. They are sufficiently stiff enough to provide structure for the finger and to produce normal force for gripping, at the same time soft enough to require a reasonable size motor for actuation. Their dimensions and spring constants are included in Table 2.1. The spring constant, k, is calculated from the dimensions and the number of active coils, or the number of coils which separate the two

27 18 finger segment links, not the entire number of coils in the compression springs. The spring constants are calculated according to the equation 4 Gd k =, where G is the shear 3 8nD modulus of steel which is approximately pounds per square inch. In Table 2.1, the individual joints are labeled with an i j convention where i refers to the finger and j refers to the joint. The thumb corresponds to i = 1 and the index finger to i = 2 and so forth. The joint closest to the fingertip of each finger corresponds to j = 1, the middle joint to j = 2, and the joint closest to the hand corresponds to j = 3. Table 2.1 Spring properties

28 Design of the actuation system The design of the hand mechanism calls for the joints to be actuated by cable and conduit sets. Since the hand was designed to be incorporated into either a robotic or prosthetic system, the most likely method of doing the work required for actuation is using electric motors. The actuation system, Figure 2.7, links the motors to the cable conduit sets and the individual motors to each other, in such a way that individual cables can be displaced while holding the conduit in place. Figure 2.7 Actuation system Once the hand mechanism itself was developed, an actuation system was necessary to explore its capabilities and refine its design. For the research and

29 20 development stage of the hand prototype, a bench-top actuation system serves well for experimentation and further refinement. This bench-top system is designed to be simple, cost effective, and configurable. It is not optimized to fit in a small, ergonomic package similar to what would be necessary for integration into a robotics or prosthetics application. Included in the actuation system are the cables that actuate each finger segment and the conduit that houses and routes the cables. As tension is applied to the free end of the cable, an equal and opposite force must be applied to the same end of the corresponding conduit to prevent movement of the entire hand. This force is transmitted down the length of the conduit to the other end of the conduit and to the finger segment where it is constrained. The equal and opposite force restricts the tension in a cable to actuating only the intended finger segment. The actuation system also contains five stepper motors which each turn three pulleys; each pulley actuates one cable each. The rest of the actuation system consists of assorted hardware to connect three pulleys to each motor and to connect one cable to each pulley (Figure 2.8), and a fixture to hold all of the motors and pulleys (Figure 2.9). Finally, 15 cable adjusters individually adjust the pretension in each cable in order for all three cables in one finger to begin to displace all the finger segments simultaneously. The cable adjusters screw in or out of the fixture, in effect lengthening or shortening the conduit so the that the tension in a each cable can be adjusted with respect to the tension of the other two cables for that finger. In this way the finger segments for a given finger can all begin to move at a certain orientation of the stepper motor and can reach full displacement at the same time at a second orientation of the motor. The cables are 3/64

30 21 of an inch diameter steel wire rope or aircraft cable. At first 1/16 of an inch diameter cable was used but it was determined that it was too stiff to wrap around such small diameter pulleys. Switching to 3/64 diameter cable solves this problem well. Currently enough cable and conduit was included to locate the hand approximately three feet away from the rest of the actuation system. To reduce cost and complexity for research and development stage, the joints are coupled to reduce the DOF s of the actuation system for the hand. Joint coupling also reduces the number of motors that would be potentially required. This is a big advantage when considering that the hand mechanism is meant to be part of a robotic or prosthetic system. The coupling is achieved by actuating all three DOF s for each finger with one stepper motor that rotates three pulleys. The circumferences of the pulleys were varied with respect to each other according to the maximum displacement of the cable necessary to draw the corresponding finger segment to its maximum defection. This way, the three finger segments in a given finger would be at the same percentage of full deflection at all times. All three segments would begin to move simultaneously. They would also reach their maximum displacement simultaneously at one orientation of the motor. Coupling the joints in this manner provides a motion that is similar to natural human flexion and extension of the fingers because human DIP and PIP joints move in a coupled manner as well. The pulley system assembly is shown in Figure 2.8. This system incorporates timing pulleys to wrap up the cables. Timing are simple, cogged pulleys that are typically available in various sizes, but are generally relatively small and are designed to be used with cogged belts that transfer power between two or more axes without slip.

31 22 This application is much different than the intended one but the pulleys perform well and were a convenient choice. This is because they were readily available in the correct sizes, were economical, and required little modification to work with the other components. Figure 2.8 Pulley assembly with shaft extension, coupler, pulleys, bearing, and nut To turn three pulleys with each motor, it was necessary to create an assembly to lengthen the motor shafts. This was done by producing a shaft that was the same diameter as the motor shaft, had a flat notch for a setscrew and was threaded at the other end. This was connected to the shaft of the stepper motor by use of a coupler with two setscrews to engage both the motor shaft and the shaft extension. This allowed room for three pulleys, a bearing, and sufficient threads still exposed to hold a nut on top while

32 23 leaving room in between the pulleys to pinch three cables. The pulleys are kept from slipping around the surface of the shaft by use of a setscrew that sets on to same notch as the coupler. The cables are gripped by the pulleys by means of a notch cut into the hubs of the pulleys. A cable retainer, which is essentially a ring, then fits around the hub to keep the cable from slipping out of the notch. A space is then left in between each pulley and the pulley above it or in the case of the upper most pulley the space is left in between the pulley and the bearing above it. This space allows the cable to be wrapped around the shaft after it is run through the notch so that when the nut on top is tightened all three cables are pinched and the cables are locked firmly to the pulleys. A bench top fixture was constructed to hold several components including the stepper motors, pulley assemblies and cable adjusters and performs several other functions necessary to actuate the hand. The fixture also houses the stepper motor drives and two switches all of that will be discussed in more detail in chapter three. See Figure 2.9. This fixture may not be a practical enough means of housing all the components to be suitable for robotics or prosthetics applications, but serves its purpose well for the developmental stage. One of the other functions fulfilled by the fixture is to give the conduits something to push against as the cables are tensioned allowing the actuation of the fingers to occur. Recall, that as a finger is actuated the conduits push on one finger segment link as the cables pull on another. The fixture also has a plate that can be slid horizontally and then held in place with two bolts, adjusting its distance from the five stepper motors. This plate provides support to the free end of the motor shafts by providing a surface to support the normal forces that are transmitted through the bearings on the ends of the

33 24 shafts as the cables are tensioned. Transmitting forces through the bearing to the motor fixture lessens the bending moment being applied to the end of the motor shaft by changing the loading conditions from those similar to a cantilever beam to those of a simply supported beam. Figure 2.9 Fixture assembly with motors, motor drives, and pulley assemblies

34 Finger deflection The change in shape of the fingers as they are deflected is important to understand when planning motions that will be executed by a control system. The deformed shape and fingertip trajectory determine where the finger will come into contact with an object as it is grasped. Fingertip orientation as the finger is flexed towards full flexion is also important to understand. This is so that the points on the fingers that contact an object as it is grasped can be predicted. A series of experiments to measure these characteristics is presented below (Figure 2.10). Figure 2.10 Trajectory and range of motion experimental setup

35 26 Each finger was measured individually. They were removed from the rest of the hand and fitted with an end effecter (needle) at the fingertip. The base of the finger was then fixed in a clamp. The initial position and the orientation of the end effecter were recorded on paper behind the finger. The finger was initially positioned at full extension and then was stepped in small increments, 20 steps apart (see Chapter 3), until full flexion was reached. Measurements were taken to determine the position of the end effecter in the flexion / extension plane and orientation of the end effecter in the same plane (Figure 2.11). The joint angles of each segment were measured at full extension and full flexion to determine range of motion (Table 2.2). Joint angles are defined as the angle of deflection from one end of the spring to the other. They are determined by the orientation of one finger segment link with respect to the previous one. These experiments point out that range of motion of the fingers is one area where there is room for improvement in the design of the hand. It is suggested that the total range of motion for the fingers of the human hand is 215 degrees for the thumb and between 270 and 300 degrees for the index, middle, ring, and little fingers (Peña Pitarch et al, 2005). Because of this deficiency, the hand may have trouble grasping small objects in certain grasp types, for instance grabbing a pencil in a cylindrical grasp. Results from these experiments, specifically the maximum range of motion, were used to develop the motion control program discussed in Chapter 4. The maximum rotation of the motor must be observed so that the motors do not try to move the fingers past their maximum deflection point. These experiments point out that range of motion of the fingers is one area where there is room for improvement in the design of the hand. It is suggested that the total

36 27 range of motion for the fingers of the human hand is 215 degrees for the thumb and between 270 and 300 degrees for the index, middle, ring, and little fingers (Peña Pitarch et al, 2005). Because of this deficiency, the hand may have trouble grasping small objects in certain grasp types, for instance grabbing a pencil in a cylindrical grasp. Table 2.2 Joint range of motion Finger Thumb Index Middle Ring Little Joint Minimum joint angle (degrees) Maximum joint angle (degrees) Range of motion (degrees) 1-1, interphangeal , metacarpophalangeal , carpometacarpal finger range of motion 2-1, distal interphalangeal , proximal interphalangeal , metacarpophalangeal finger range of motion 3-1, distal interphalangeal , proximal interphalangeal , metacarpophalangeal finger range of motion 4-1, distal interphalangeal , proximal interphalangeal , metacarpophalangeal finger range of motion 5-1, distal interphalangeal , proximal interphalangeal , metacarpophalangeal finger range of motion 167 Maximum roation of motor (degrees) Results from these experiments, specifically the maximum range of motion, were used to develop the motion control program discussed in Chapter 4. The maximum rotation of the motor must be observed so that the motors do not try to move the fingers past their maximum deflection point.

37 28 Thumb Index Y Position (mm) Y Position (mm) X Position (mm) X Position (mm) Middle Ring Y Position (mm) Y Position (mm) X Position (mm) X Position (mm) Little 70 Origin and fingertip location and orientation on finger. 60 Y Position (mm) X Position (mm) Figure 2.11 Fingertip trajectory and orientation

38 Aesthetics The need to have a realistic, esthetically pleasing device is especially necessary in the field of prosthetics. For many prosthesis users, the aesthetics of a device is just as important as the functionality. The ideal replacement for the human hand is one that perfectly mimics not only the function of the human hand but also the appearance including size, shape, weight, texture, color, and movement. The user of a prosthetic device has to be very comfortable using it and the ideal comfort level naturally includes appearance among other things (Kyberd et al, 2003). To help achieve this ideal, this hand was intentionally designed for use with a cosmetic glove covering it. The glove resembles a human hand and forearm, and this dictated the dimensions of the majority of the components used in the construction of the hand. The diameters of the finger segment links and of the springs were chosen so that they would closely match the inner dimensions of the cosmetic glove. The lengths of springs were carefully chosen so that middle of the spring lengthwise, and therefore the middle of the curve of the deflected finger segment, would coincide with the location of the knuckles of the cosmetic glove. In its current state, improvements in the hand s structure could still be made to achieve a closer resemblance to the shape of a human hand, especially in the case of the hand body. To achieve the ideal shape for an entire hand assembly would necessitate a close focus on human hand anthropometry when designing all the components. Results for the developmental stage of this more realistic mechanical hand have been satisfactory. Figure 2.12 and Figure 2.13 compare, respectively, the palm side and the back of the hand of the mechanical hand with a human hand. Each photo shows the

39 30 human hand on the right, the artificial hand on the left. Although there is some difference in color between the human and cosmetic glove, which is available in many skin tones, the shape of artificial hand does mimic the human hand well, especially considering that improvements could be made with a more anthropometrically driven design. Figure 2.12 Artificial and human hand, palm Figure 2.13 Artificial and human hand, back of hand.

40 31 CHAPTER III ELECTRICAL DESIGN This chapter discusses the design of the electrical components of the hand and the requirements the design must satisfy. The requirements considered include the amount of tension that must be produced in the cables to fully displace each finger and produce a sufficient gripping force, the length of cable displacement necessary, and the resolution of movement necessary to produce precise finger movements. Component selection is critical to satisfying these requirements and defines the specific challenges that must be overcome to achieve good performance of the hand. Small DC motors are the best choice to actuate the hand. This decision leaves two options, servomotors or stepper motors. The advantages and disadvantages of each as well as the reasoning for choosing stepper motors to actuate the hand are discussed below. The operating principles and strategies of stepper motors, five of which form the basis of the electrical system, are also discussed. Understanding how a stepper motor works is background information but it is also essential to selecting the proper components, which ultimately determines how the hand will function. Finally, the entire electrical setup that actuates the hand is detailed. The actual performance of the entire system is also discussed. 3.1 Performance requirements The motors that will provide the best solution to actuate the hand, that is provide the best torque, speed, acceleration, and positional accuracy characteristics, depend on

41 32 what kind of operating conditions they will experience. Therefore, when selecting motors and other required electronic components to power the hand, it is necessary to consider everything that will be required of them when displacing the fingers or grasping an object. The major requirements considered include the amount of torque required to fully deflect the fingers and exert gripping force on an object, the speed at which a finger should move in order to mimic human motion, and positional accuracy of the motor needed for fine manipulation. The first requirement to consider is the amount of force it takes to displace the fingers to full flexion and how much torque is required of the motors to exert this much force on the cables. This is the minimum force required to produce the closed hand posture. Additional force will need to be generated in order to produce significant normal force on the surface of a grasped object. It is also just as important to consider the pulley radii that will be used with the motors, as this will give a direct relationship between the force necessary to pull the cables and the torque required to produce this force. The smaller the radius used for each pulley the less torque required to generate the same amount of tension in the cable and therefore the smaller motor required. On the other hand, there is a limit to how small a pulley can be. The outside diameter, the diameter at which the cable will contact the pulley, must be at least some minimum amount greater than the bore diameter, or the diameter of the hole in the center of the pulley that will fit the motor shaft. Therefore, the torque rating of the motor cannot be considered independently, but must be considered at the same time as size of the pulley and the diameter of the motor shaft. Table 3.1 lists the force necessary to displace each joint to full displacement and the total force necessary to displace each finger. Also listed

42 33 is the required torque for the pulley sizes used. The joints are labeled according to the same i, j notation as described in Chapter 2. Table 3.1 Force and torque requirements for each joint The sizes of the pulley used, more specifically the radii are based on several considerations. First is the smallest pulley outside diameter possible to use with the shaft diameter of the motor. Also it is necessary to have a suitable relationship between the three pulley radii for one given finger so that the three finger segments will reach full displacement at the same time. The third consideration is the standard size pulleys readily available. The pulley radii for each joint are also listed in Table 3.1. Another consideration to take into account is the maximum and minimum change in orientation that the motor will be expected to make in one movement. Since the average displacement required for each joint is less the one inch of cable (listed in Table 3.1) and the displacement for any individual finger joint is less than the circumference of

43 34 the corresponding pulley, no motor will be required to rotate the output shaft of the gearbox an entire 360. Therefore the movements that require the highest angular velocities are very short in duration. Because of this, a motor that can accelerate and decelerate relatively quickly is required for quick movements, whereas, grasping and manipulation many times requires very fine movements. Therefore a motor that can be precisely positioned is also necessary. If no other dedicated device, such as an electrically triggered brake, is used to maintain cable tension in static loading situations, an additional requirement of the actuation system is that the motors provide a means to hold the position of the hand stably. 3.2 Stepper motors versus servomotors Electric motors are the obvious choice to actuate the fingers of the hand when considering either robotics or prosthetics applications. No other actuation method, such as pneumatic or hydraulic, is as reliable or as easy to implement. Production of a marketable product for either a prosthetics or robotics application will likely require a device that can run on battery power. This naturally leads to the choice of using motors that run on direct current (DC) rather than motors that are powered by alternating current (AC) since power from the battery is already DC. Considering the requirement for a DC motor, and torque and speed requirements there are two basic motor categories that could actuate the hand, servomotors and stepper motors. This is a critical decision as it has a large impact on overall performance and greatly dictates many of the other system components to be used.

44 35 Servomotors in a size range suitable for this application are typically DC brush or brushless motors. They use a position sensor, usually potentiometer based with an analogue signal or an optical encoder, to provide position feedback for closed loop control. Using an electronic controller, accurate positioning can be achieved similar to a stepper motor, but a feedback signal is a necessity. Servomotors generally have a higher rpm range than stepper motors and produce more torque at higher velocities than at lower ones. Stepper motors are electric DC motors with no commutators. In most other electric motors the commutators switch the electromagnetic poles in the motor so that the rotor is constantly made to turn. In the case of stepper motors all the commutation is handled by external circuitry generally referred to as a stepper motor drive. The drive will energize the correct magnetic pole or poles to advance the motor to the next step, hence the term stepper motor. The windings are then energized in the proper sequence and rate to make the motor rotate the desired direction and at the desired rate. A stepper motor with the correct power supply and drive can be rotated at relatively high speeds, although generally a stepper motor cannot easily be made to rotate as fast as a servomotor. But stepper motors can accelerate or decelerate at relatively high rates and can be turned to a precise position and then hold that position. One advantage that stepper motors have is that in many applications they can be controlled with open loop control, which doesn t require any feedback from an encoder. The encoder reports the position of the motor to the controller so that errors can be corrected. As long as the stepper motor doesn t slip or fail to advance to the next step when the drive commands it to, then the stepper motor can be controlled in an open loop. This is

45 36 possible in applications where the loads and accelerations placed on the motor do not exceed its maximum torque (Jones, 2004). 3.3 Stepper motor theory and operation This section discusses the three basic types of stepper motors, permanent magnet, variable reluctance, and hybrid, which is a cross between the previous two, advantages and disadvantages of each, strategies for achieving the greatest performance from a stepper motor, and the physics behind these concepts. The basic difference has to do with the construction of their rotors and the arrangement of their windings. The permanent magnet stepping motor relies on the electromagnetic interaction of an energized stator winding and a permanent magnet rotor. One thing that differs between a permanent magnet or a hybrid stepping motor and a variable reluctance stepping motor is that fact that the permanent magnet and variable reluctance motors maintain a fraction of their holding torque even the windings are not energized. This is because the poles of the permanent magnet in either one of these motors are attracted to the stator poles. The torque present when no windings are energized is often referred to as detent torque. A permanent magnet motor can be operated constructed as either a unipolar or a bipolar motor, both of which will be explained shortly (Jones, 2004). The variable reluctance stepping motor has no permanent magnet rotor, and therefore relies on the principle of minimizing the reluctance along the path of the applied magnetic field. The stator has a magnetic core and is constructed with a stack of

46 37 steel laminations and the rotor, which has teeth and slots, is made of soft steel that is not magnetized (National Instruments, 2005). Because the variable reluctance stepper motor does not have a magnet rotor it has no detent torque and will rotate freely when no current is supplied to the windings. Hybrid motors combine several characteristics of both variable reluctance motors and permanent magnet motors. They combine a permanent magnet rotor and multi toothed stator poles made of soft steel and are very similar to a permanent magnet motor in terms of control (Jones, 2004). The stator of a hybrid motor is very similar to a variable reluctance motor except for one aspect. In a variable reluctance motor, only one coil of each phase is wound around each pole. Usually in hybrid motors, which have what is known as a bifilar connection, there are two coils wound around each pole one from two different phases. Torque is then created by the magnetic interaction of the permanent magnet rotor and the stator (National Instruments, 2005). Permanent magnet and hybrid stepper motors can either be unipolar or bipolar. The difference is defined by the arrangement of their windings and determines the kind of drive that must be used to power them. A unipolar motor has two windings or phases. The windings have a center tap which is usually connected to the positive supply. Then one or the other of the other two ends is grounded (Figure 3.1). This reverses the current flow though the winding and the direction of the magnetic field produced by the winding. Unipolar motors can be operated by energizing either half of one winding at a time, or half of both windings at a time (Jones, 2004). The advantage with a unipolar motor is that it requires less sophisticated drive circuitry. The output current from the drive is always in one direction. The disadvantage with comparison to a bipolar motor is that at any one

47 38 time, at best only fifty percent of the total windings in the motor are being used. This results in less torque for a given size motor with comparison to a bipolar motor (Lin Engineering, 2005). Figure 3.1 Unipolar stepper motor Jones, D., Control of Stepping Motors, (2004), < (19 November 2004) A bipolar motor is constructed essentially the same as a unipolar motor except that the windings are simpler. There are no center taps (Figure 3.2). The simpler winding circuitry leads to a simpler overall motor design. Although the motor itself is

48 39 simpler, it requires more advanced control circuitry because it is necessary to reverse the direction of the current through the windings (Jones, 2004). The advantage a bipolar motor has is that one hundred percent of its windings can be used at any given time. Also more torque is produced. The amps per phase equals 1 2 of the amps per coil when two coils are connected in series, as in a bipolar motor. If N represents the number of turns per coil then N I is proportional to the torque in a unipolar motor, then 2N ( 1 2 ) I, or 2NI, is proportional to the torque in a bipolar motor, approximately 40% more then a unipolar motor (Lin Engineering, 2005). Figure 3.2 Bipolar stepper motor Source: Jones, D., Control of Stepping Motors, (2004), < (19 November 2004)

49 40 Although a typical hybrid stepping motors, due to the design of their rotors and windings, have 200 steps per revolution, drive electronics that can utilize more advanced electromagnetic control strategies can achieve much finer angular resolution. A motor with 200 steps per revolution is said to have a step angle of 360 /200 or 1.8. When a drives divide a step in two, this is commonly referred to half stepping. Anything finer than half stepping is commonly referred to as micro stepping. As long as none of the magnetic circuit is saturated, powering both motor windings simultaneously will produce a torque versus position curve that is the sum of the torque versus position curves for the two motor windings taken independently (Figure 3.3). The two curves will be S radians out of phase for a two-winding permanent magnet or hybrid motor, where S is the step angle. If the currents in the two windings are equal, the peaks and valleys of the sum will be S/2 radians from the peaks of the single winding curves. This is the key to half-stepping. The two-winding holding torque is the maximum of the combined torque curve when both windings are carrying their maximum current. For common two-winding permanent magnet or hybrid stepping motors, the two-winding holding torque is, again, 2 times the single winding holding torque. This assumes that the magnetic circuit is not saturated and that the torque versus position curves are ideal sinusoids. Micro stepping is an extension of this idea that uses two different current levels through the two motor windings as in Figure 3.4. Common stepper motor drives divide steps in to half steps, on quarter steps, and one eight steps. Some stepper motor drives are capable of even finer precisions.

50 41 Figure 3.3 Torque verses position for a single winding and two winding motors Source: Jones, D., Control of Stepping Motors, (2004), < (19 November 2004) Figure 3.4 Torque versus position for two windings using micro stepping Source: Jones, D., Control of Stepping Motors, (2004), < (19 November 2004) For a two-winding variable reluctance or permanent magnet motor, assuming nonsaturating magnetic circuits, and making the same assumptions as before, the following two formulas give the key characteristics of the composite torque curve: 2 2 h a b 1 = + and x tan ( b a) S =. Where: a equals the torque applied by the π 2 winding with equilibrium at 0 radians, b equals the torque applied by the winding with equilibrium at S radians, h equals holding torque of composite curve, x equals

51 42 equilibrium position, in radians, and S again equals step angle, in radians. With no saturation, the torques a and b are directly proportional to the currents through the corresponding windings. The two currents are then varied and it is possible to achieve 1/4, 1/8, or smaller step sizes (Jones, 2004). An important consideration in operating stepper motors at high-speed is the effect of the inductance of the motor windings. Rise and fall time of the current through the windings is a factor of inductance of the motor winding. Ideally the current versus time would be a square-wave but the inductance of the winding causes it to be an exponential, as illustrated in Figure 3.5 Figure 3.5 Current through winding versus time Source: Jones, D., Control of Stepping Motors, (2004), < (19 November 2004) The exact characteristics of the curve of current through each winding versus time depend as much on the drive circuitry as they do on the motor. These time constants of these exponentials can easily differ. The rise time is a condition of the drive voltage and drive circuitry, while the fall time is determined by the circuitry used to dissipate the stored energy in the motor winding.

52 43 At relatively low step rates, the rise and fall times have much less effect on the motor s torque than at high step rates, as shown in Figure 3.6. This is because the ratios of rise and fall time to the duration of time a winding is energized is much lower at slow step rates compared to higher ones. At low step rates the winding is conducting full current for a greater percentage of the time that the winding is energized. This leads to greater running torque at low speeds. Figure 3.6 Torque versus speed Source: Jones, D., Control of Stepping Motors, (2004), < (19 November 2004) The motor's maximum speed is the speed at which the available torque goes to zero. A curve of torque versus speed for a typical motor and control system can usually be approximated by a horizontal line at low step rates and a line with negative slope going to zero over the range of higher step rates. The cutoff speed is defined as the step rate at which these two regions of the curve meet. A definite cutoff speed is rare, therefore, statements of a motor's cutoff speed are approximate. The rise and fall times of

53 44 the current through the motor windings occupy a relatively small percent of each step when the motor is operating at rates less than its cutoff speed. While at the cutoff speed, the step duration is comparable to the sum of the rise and fall times. The exact torque versus speed curve depends on the rise and fall times in the motor windings, and these depend on the motor control system as well as the motor. Therefore, the control system for a motor also has a large effect on the cutoff speed and maximum speed, not just the motor itself (Jones, 2004). One strategy to improve a motor s cutoff speed, maximum speed, and high-speed torque is current limiting. Increasing the voltage applied to the windings increases the current through the windings in a simple V = I R relationship, where V is the voltage, I is the current, and R is the resistance of the windings. Increasing the voltage applied to the winding to a level which results in a current that is significantly higher than the rated current results in much quicker rise and fall time. Unfortunately current levels this high result in damage to the motor; usually the thermal breakdown of the insulating material in the motor windings. The idea behind current limiting is to use a voltage significantly higher than necessary to achieve the rated current, but to use advanced circuitry to drop the voltage applied to the windings, once the rated current level is reached. In this way, the current versus time curve comes much closer to resembling a square wave, as in Figure 3.7. The advanced circuitry techniques used to achieve current limiting include resistive current limiters, linear current limiters, open loop current limiters, one-shot feedback current limiters, or hysteresis feedback current limiting. Current limiting technology is also required to achieve micro stepping since at times during micro stepping one of the motor windings is run at less than the rated current (Jones, 2004).

54 45 Figure 3.7 Current versus time with and without current limiting Source: Jones, D., Control of Stepping Motors, (2004), < (19 November 2004) 3.4 Stepper motor, drive, and power supply setup For the research and development stage of the hand mechanism, a system that is easily reconfigurable and expandable has obvious advantages. The overall system that converts electrical power to mechanical power consists of stepper motors, stepper motor drives, and a power supply. This section describes how all the separate components that were selected based on the performance requirements and operating principles discussed above are combined in one system. It also discusses the performance of the system. Five NEMA Size-17, bi-polar, hybrid, 1.8, DC stepper motors equipped with 3.6:1 gear reduction via an offset spur gear provide mechanical power for the hand. Five RMS Technologies, R208 drives power the motors. They have micro stepping and current limiting technologies. Two variable voltages linear power supplies from BK Precision, which run on 110 VAC and provide an output of 0 to 15 VDC at 40 Amps, provide the electrical power. They are connected in series, providing variable voltage over the input range of the drives, 12 to 24 Volts.

55 46 This system provides excellent performance characteristics suitable for actuating the hand mechanism. The angular resolution of the motors allows for very precise control of the cable displacement, and therefore, hand posture. With the 3.6 to 1 gear reduction, this setup is capable of step sizes as small as 0.5 without micro stepping, or using the finest level of micro stepping, one eighth stepping, the drives are capable of rotations as small as Depending on the pulley size for a given joint, these rotations correspond to cable displacements as small as approximately mm and mm, respectively. This particular combination of drives, motors, and gearing produces running torque of 1.06 N-m (150 oz-in) at low speeds and 0.81 N-m (115 oz-in) at 2 revolutions per second, as reported by the manufacturer (Figure 3.8). Figure 3.8 Manufacturer s torque rating (oz-in) of motor versus speed (revolutions/second) Source: Lin Engineering, Lin Engineering FAQ, (2005), < (29 March 2005)

56 47 These torque and velocity figures refer to torque and velocity after gear reduction and not torque and velocity of the motor itself. Holding torque, or the torque the motor provides to keep the hand static, is critical to maintaining a grasp. The holding torque produced when the maximum current is supplied to the windings while the motor is stationary would be approximately equal to the maximum running torque, but the friction associated with the gear reduction increases this amount. The amount of torque produced provides tension on the cables that is quite sufficient for posturing the hand and providing gripping force. The mechanical aspects of the hand have dictated the design of the electrical system that powers the hand. The most critical aspects were the amount of tension required in the cables to displace the fingers fully and the corresponding displacement of cable. Due to these mechanical aspects of the hand, five stepper motors with gear reduction were determined to be the most suitable devises for converting electrical power into mechanical movement. With suitable drives, which are capable of current limiting and micro stepping, the stepper motors provide sufficient torque over the entire angular velocity range, especially at the low end, and at the same time are capable of extremely fine movements. Since stepper motors are relatively inexpensive it is wise to plan to size a motor large enough to provide torque two or three times greater than the maximum expected required torque. Of course, as the process of optimizing the design for a prosthetics or robotics application is begun, the use of oversized motors may no longer be practical. Once the appropriate stepper motors are selected, this decision determines the other necessary electrical components: stepper motor drives, and a power supply.

57 48 CHAPTER IV MOTION CONTROL This chapter discusses the computerized motion control for the hand, why such control necessary, various alternatives for achieving it, and specifics of the method used. A user interface based on a hierarchical control scheme that simulates how the hand could be controlled in the real world is discussed. New software, which is discussed in the chapter, was developed to implement this control scheme. Also, further advancements necessary to achieve a useable level of controllability of the hand are discussed. 4.1 Motion control setup To achieve coordinated motion of five DOF s which mimics human motion, requires computerized control. At a lower level of control, simple finger motions, computerized control is necessary to manage simpler requirements such as maintaining the joint limits of the hand and defining a single move, such as number of steps, direction, maximum velocity, rate of acceleration, rate of deceleration, etc. At a higher level of control, computerized control is also necessary for inter-coordination of five independent DOF s. Computerized control is especially necessary to implement a hierarchical control scheme which can simplify the necessary input from a prosthesis user to a level where managing a high number of DOF s is comfortable.

58 49 The term hierarchical control scheme refers to one in which it is possible to control a high number of degrees of freedom while the user only inputs a smaller number of DOF s, in prosthetic control typically one or two. For instance a user may only signal a higher level command such as close grasp and the control scheme would generate corresponding lower level commands such as move the first axis 30 counter-clockwise, move the second axis 45 counter-clockwise, etc. Additional information could also be collected be external sensors in the hand to help make decisions automatically, such as stop tightening the grasp when an object comes into contact with one or more fingers. Since stepper motors have no commutators like other DC motors, all the commutation, or continuous change in the electromagnetic field to turn the rotor, must be handled by external electronics. The device that is responsible for this is referred to as a drive. At the very least the stepper motor drives require two inputs signals to operate, direction and pulse rate. The direction signal is a simple on - off signal of plus or minus 5V to indicate the direction of travel, while the pulse rate signal is a sine wave, the frequency of which indicates the pulses per second which is directly proportional to the number of steps per second. The drives then energize the appropriate windings of the stepper motor at the appropriate times. When considering the most basic aspect of control for this application, sending five direction and pulse signals to five stepper motors drives, the are several categories of electronics that could possibly perform this task as well as more complex tasks. These categories include electronics specifically designed for this task such as PCI or ISA bus motion controller cards that reside in a host PC, a range of stand alone motion controllers that are an analogue of the PCI or ISA bus versions but operate removed from the PC, as

59 50 well as more general purpose electronics which could be adapted to perform the task such as Programmable Logic Controllers (PLC s), PC-104 based electronics, or a custom designed electronics board. For the current state of the project, research and development, a system that is modular and easily configurable, both in terms of hardware and software, greatly aids in experimentation and refinement by allowing the easy addition of new hardware components or programming. After examining all the alternatives, a control system based on a PCI bus motion controller was determined to be the most practical solution. It is simple to implement, designed specifically to perform all of the necessary control tasks, easily configurable in terms of software and hardware, and relatively cost effective. The motion controller is a DMC-1850 five-axis motion controller that resides in the PCI Bus of a standard desktop PC. The number of axes refers to the number of DOF s or motors that it can control. It has its own microprocessor and memory and performs all of the motion commands internally without using the computers resources (Figure 4.1). Figure 4.1 Actuation and control setup

60 51 The stepper motor drives used are RMS Technologies, R208 bipolar drives. They have current limiting technology, so they can run on power anywhere from 12 to 24 VDC. They also have a peak current which is adjustable from.35 to 2 Amps peak, to suit the particular motor and application. They are capable of step sizes of full, half, and 1/4 and 1/8 micro-stepping. Other features include optically isolated step, direction, and enable/disable inputs, and a current cutback feature that can be disabled. When enabled, the current cutback feature reduces the holding current to 23% of peak current to reduce temperature buildup and energy consumption. They have a 9-pin input connection, the other six that have not been mentioned previously include: main power and ground, a 5V power and ground for the logic circuits, and two inputs which can be enabled and disabled in four different combinations to select the step size. In this way the controller can control all the functions of the drives. A detailed wiring diagram is included in Figure 4.2. The interconnect module is basically an extension of the motion controller and provides terminals to handle all of the input/output for the motion controller. This includes committed I/O for each axis for pulse rate, direction, and an enable signals (which activates each drive), along with several digital and analog inputs and outputs, which allows for a great deal of expandability ideal for this stage of development. The particular motion controller used actually requires two interconnect modules because each module can handle up to four axes. This setup has five motors. The interconnect module terminals included in the figure are only the ones needed for to control the functions of the drives, there are several more which can be used to add a variety of hardware.

61 52 Power Supply Interconnect VDC ground signal ground + 5VDC A axis direction A axis pulse output B axis direction output B axis pulse output C axis direction C axis pulse output D axis direction D axis pulse output V 2. step 3. step 4. enable 5. direction 6. ground 7. logic ground 8. +5V 9. PWM B B A A Drive, A A axis enable B axis enable C axis enable D axis enable output 1 output 2 output 3 output 4 output 5 output 6 output 7 output V 2. step 3. step 4. enable 5. direction 6. ground 7. logic ground 8. +5V 9. PWM B B A A Drive, B Interconnect E axis direction E axis pulse output E axis enable V 2. step 3. step 4. enable 5. direction 6. ground 7. logic ground 8. +5V 9. PWM B B A A Drive, C output 9 66 output V 2. step 3. step 4. enable 5. direction 6. ground 7. logic ground 8. +5V 9. PWM B B A A Drive, D V 2. step 3. step 4. enable 5. direction 6. ground 7. logic ground 8. +5V 9. PWM B B A A Drive, E Figure 4.2 Wiring diagram

62 53 Explicit commands can be sent for execution by the motion controller from an application running on the PC that is designed specifically for use with this motion controller. This of course, is a very basic way of controlling the hand and is only useful for developmental purposes. This method does not allow for the possibility of executing a coordinated series of motions. executable running on the PC. The motion controller can also interact with an In this way virtually any level of logic can be incorporated in the control of the actuation system. The highest-level commands are generated in the executable running on the PC. Explicit commands, such as specific relative position moves with a defined acceleration, deceleration, and velocity, can then be sent to the motion controller card. The motion controller then converts these commands to the appropriate pulse rate and direction signals for each axis and then sends them to the motor drives via the interconnect module. The motor drives then use these pulse rate and direction signals to energize the windings of each stepper motor in the appropriate sequence to complete the required motion. The inputs can be used to read nearly any variety of external sensors, which can then be monitored by the motion controller. The executable can then query the motion controller for the state of these inputs. In this way, the external sensors can be used to trigger events that happen in the executable. For example, as part of a hierarchical control scheme, pressure sensors could be incorporated into the palm and fingertips of the hand. Myoelectric signals from the user could be monitored until the appropriate signal was detected to indicate that the user intends to close the grasp of the hand. The hand would begin to close until a change in the signal from the pressure sensors indicated that contact with an object had been

63 54 achieved causing the hand to hold its position after the appropriate pressure has been applied. The system also includes a main power switch in the positive wire (Figure 4.2) in between the power supply and the motor drives that is installed on the motor fixture, a 20 Amp fuse in between the power supply and the switch, and an abort switch. The abort switch is wired to a designated input, that when activated terminates any motions currently being executed and prevents other motions from being executed until the switch is reset. One addition to the system, which would not only greatly improve performance of the current system, but would also be a necessary element of the control system in any form suitable for a consumer product, would be encoders or some other device which could provide a position feedback signal in order to correct positional errors. Since when stepper motors move they move a commanded number of steps, which corresponds to a certain angular position, stepper motors work very well in some applications that do not include positional feedback. This is true when the inertia of the load and attempted accelerations do not exceed the available torque of the motor. If the load on the motor does exceed the available torque, the stepper motor will slip or fail to advance to the next step. In this particular application the end effectors purposely collide with a grasped object, and therefore the maximum range of motion are, in effect, different for every object the hand grasps. A positional feedback signal could detect a slip by the stepper motors, which could signal contact with a grasped object or other cases in which the motor slips. Without this feedback, once a slip occurs, all knowledge of position is lost.

64 Programming This section discusses new software that was developed to implement a hierarchical control scheme. A generic method for programming the hardware used and basic capabilities are discussed first. A software development kit is provided with the motion controller card. It includes, dynamic link libraries, (DLL s) which can be used with any Windows programming environment that interfaces with DLL s. The basic programming model includes six steps, described as follows. The first step is to the declare functions. The second step is to start a communication session between the executable and the controller. The third step is to download a program. If necessary an entire predefined program can be downloaded from the computer to the controller and then executed on command. Step four is to send live commands including axis motions, input activation, and input querying. The last step after the rest of the program is complete is to close communication. The software developed to control the hand follows these basic steps. One of the most difficult aspects of developing a hand mechanism that could potentially be used as the major component of a prosthetic system with a large number of DOF s is a user interface and control scheme. Many currently available prosthetics use myoelectric signals from the residual muscles to control grasping. However, to consciously control more than one or two DOF s individually is too mentally complex for the user and would make the device too cumbersome to use. Control of a large number of DOF s is impossible unless some sort of control scheme can be developed to simplify the commands the user must provide and coordinate many DOF s.

65 56 To demonstrate the capabilities of the hand, a control scheme was developed based on a control scheme developed for a different application. Knutson et al (2004) developed a control scheme simulating state activation of a neuroprosthesis using two myoelectric signals, from the wrist extensor and flexor muscles. Sensors were implanted in patients wrist flexor and extensor muscles, and then both myoelectric signals were monitored. The flexor myoelectric signal was then used to represent an x-coordinate; the extensor myoelectric signal was used to represent a y-coordinate, which together corresponded to a location on a state activation chart (Figure 4.3). Figure 4.3 State activation chart Source: Knutson, J., Hoyen, H., Kilgore, K., Peckham P., (2004), Simulated Neuroprosthesis State Activation and Hand-Position Control Using Myoelectric Signals from Wrist Muscles, Journal of Rehabilitation Research & Development, Vol. 41, Issue 3B, P461.

66 57 The signal space is divided into four regions, Hold, Open, Close, and Change Grasp Pattern. When the Hold command is active, which corresponds to both sets of muscles being at rest, the hand mechanism stays in its current posture. When the extensor muscles are excited, the Close command is activated, which causes the hand to tighten its grasp. Likewise, when the flexor muscles are excited, the Open command is activated widening the grasp. When both sets of muscles are excited, the Change Grasp Pattern command is activated; it can be used to toggle between different grasping postures. In order to apply this control scheme as a user interface to operate the hand software was developed to use a joystick to simulate the myoelectric signals. One DOF of the joystick simulates the extensor myoelectric signal, while the other simulates the flexor myoelectric signal in the same manner as the above-mentioned neuroprosthesis. A combination of the two signals determines the location of a control point that exists on a two-dimensional state activation chart (Figure 4.3). The joystick, which contains two linear potentiometers that are each manipulated by one DOF of the joystick, is connected to the interconnect module. The potentiometers are each connected to one of the analogue inputs and the voltage across the potentiometers is read. After the program goes through a short calibration routine to find the minimum and maximum voltages for both DOF s, the program runs in a loop, constantly monitoring the voltage of across the potentiometers, until a lack of input causes the program to time out and terminate. The signals from both potentiometers are normalized and converted into Cartesian coordinates. The joystick has been physically modified so that the tension of internal springs cause the at-rest position of the joystick to

67 58 reside at the lower left corner of the range of motion, similar to the state activation chart. The command regions of the program are also divided up in a similar manner as the chart. Moving the control point out of the Hold region by varying one or both DOF s activates one of the other commands depending on which command region is entered, Close, Open, or Change Grasp Pattern. Proportional control is the ability for the user to determine how far the device opens or closes by some means. Proportional control is incorporated into the control scheme by correlating the how far the control point travels into either region before returning back towards the Hold region to how far the hand opens or closes. Moving the joystick to its maximum range of motion in a given DOF corresponds to either 100% closure or opening. Since this is an open loop system, the program monitors position by tracking the commanded direction and number of steps of each move. Joint limits are maintained by adjusting any command that calls for a motion beyond the joint limits by reducing the number of steps in the command so that the motion takes the particular DOF up to the joint limit. Subsequent motions in the direction of the reached joint limit are ignored. The Change Grasp Pattern toggles between different sets of relationships between the five DOF s, to achieve different grasp postures. If the adduction / abduction DOF were included in the fingers the hand would be capable of several kinds of grasps. However in its current configuration the hand is still sufficiently capable to perform some grasps types. It is programmed to do a pinch grasp with the thumb and forefinger and a cylindrical grasp with all five fingers.

68 59 Use of a PCI 5-axis motion controller allows for computerized motion control system that is ideal for developing the hand due to the great expandability. Software can easily be rewritten or hardware added to experiment with new control strategies for the hand. A user interface that simulates a real world hierarchical control scheme has been completed. This allows for some rudimentary control of five DOF s while only requiring two DOF s of input from the user. Ideally a level of control would be achieved that more closely approaches a human hand in terms of intuitive use and performance. It would be necessary to include encoders for positional feedback as well as additional sensors in the hand to provide feedback to indicate contact with an object, slip of a grasped object, etc.

69 60 CHAPTER V APPLICATIONS AND RESULTS Previously, in Chapter I, potential applications of the hand mentioned are mentioned, adaptation of the hand to be used as a prosthetics device and incorporation of the hand into a robotics system as an anthropometrically based gripper. In addition to these, there is a third application for which a advanced hand mechanism could be used, as a physical model to develop a cognitive model that governs the grasping capability of a digital human and verify its accuracy. Section 5.1 discusses how the hand mechanism could be used for all three applications and what would have to be done to adapt it for each. Section 5.2 discusses the results achieved thus far and how suitable the hand would be for these applications. 5.1 Applications Prosthetic devices to replace the hand can either be hook shaped or a more anthropometrically shaped hand with a range of functionality from one or two simple grasps to purely aesthetic device. Approximately 70 percent of users in the United States use a hook over a hand devise (Doshi, et al. 1998). The fact that not all users choose to use a mechanical hand prosthesis leads to the conclusion that the devices commercially available are either too expensive or are not practical enough for all users to consider them beneficial. Typically commercially available models only have three active fingers, which are usually all activated by the same motor and therefore only have one DOF.

70 61 They are only capable of a cylindrical or pinch type grasp depending on how the hand comes in to contact with the object. Integration of the hand mechanism into a prosthetic system would require two different user interfaces, a physical one and another one that allows the user to send commands to the device. The purpose of the physical interface would be to connect the devise to the user s body. The physical interface would be customized to each user and would include a socket or cup that would allow the device to join with the user s residual limb. The conventional command interface, which allows the user to interact with a prosthetic devise, uses myoelectric sensors placed on the surface of the skin. Myoelectric control has three main advantages, accuracy of command selection, intuitiveness of control, and a quick response time of the system (Englehart and Hudgins, 2003). Typically myoelectric control can be troublesome because the sensors can move or be easily placed in slightly different locations every time they are applied which causes inconsistency in the myoelectric signals, which are inherently weak and noisy (Davalli et al, 2000). Because of this, many systems require an initialization routine to calibrate the device to the changes in the signal every time the device is put on. Myoelectric control would also require additional electronics to amplify and filter the signal. This hand device, with further advancements such as additional DOF s and the appropriate sensors combined with an advanced control system based on a hierarchical control scheme could be integrated into a prosthetic system. This system would have several potential advantages over currently available devices and devices currently being researched. Since the fingers are based on a compression spring and do not contain any rigid links, complex revolute joint mechanism, or actuators, a considerable percentage of

71 62 they re volume is occupied by empty space. This provides the potential for an extremely lightweight hand if the design for the fingers can be combined with a design for the rest of the hand that is optimized to be lightweight. Since the forces to actuate the fingers are transmitted by a cable and conduit pair, there are several options for the locations of the actuators. This could possibly present a design trade off between minimization of the perceived weight of the device and containing it in the smallest package possible. In any case the actuation system would need to be made more efficient so that smaller, lighter motors could be used and the entire actuation system could be contained in a smaller, more ergonomic package. Depending on the size of the motors that are sufficiently powerful enough and therefore the torque required to actuate the hand, there could be up to three options for the location of the motors. The first option would be in the palm of the hand. For this option the motors would have to be extremely small, since there would still likely be at least five motors. This would also increase the weight of the hand mechanism and locate the center of mass of the device further away from the users shoulder, making it more cumbersome and tiring to use. The second option would take advantage of fact that the forces are transmitted through flexible cable conduit pairs and locate the motors completely off of the limb. The motors could be clustered in an ergonomic package that is worn somewhere else on the body, possibly in a pack worn around the waist. The cables could then run from the pack, under the clothes, to the hand. The advantage would be that much of the weight could be located completely off of the limb, minimizing the burden on the user. The

72 63 disadvantage would be the addition of another component to the system that the user would have to wear. If the amount of residual limb allowed for it, the third option would be a compromise between the first two options. This would be to locate the motors in the forearm or wrist space. This would locate the weight further up the user s limb reducing the perceived weight of the device compared to locating the motors in the hand itself and at the same time eliminating the need for an additional pack to be worn. No matter where the motors are located it would be beneficial to spend effort on optimizing the entire system so that it requires the smallest motors possible. The design of the force transmission system could be optimized to reduce losses due to friction. The design of the joints could be optimized so that a given deflection of the spring takes the smallest amount of tension in the cable, while at the same time creating the most contact force at the fingertips. These two design points are key to making the overall system practical because they lead to reducing the minimum size of the motors necessary, which is beneficial for several reasons. Smaller motors are easier to locate in an ergonomic package and leave more options as to their location. Smaller motors would reduce the total weight of the device, which reduces the amount of effort on the user s part to utilize it. They would also require less power, reducing the size of the batteries necessary to power the device for a given amount of time, additionally reducing weight and bulk. The hand could also be used as an anthropometrically based end effecter or gripper for a robotics system. Applications that require this sort of gripper over a more primitive pincher-like gripper are those that will model their gripping strategies on the strategies that a human employs. The applications that are likely to use an

73 64 anthropometrically based gripper are the ones in which the system would be required to perform operations that were unforeseen and varied in nature and therefore required a gripper that would be versatile and easily adaptable for many different tasks instead of a more primitive pincher-like gripper may only performs one kind of task well. For example, humanoid robots, that one day could be part of everyday life, would be interacting with the objects found in everyday life. Most of these objects, whether deliberately or not, are all to one degree or another designed to be manipulated by human hands. Therefore the most logical end effecter to manipulate these objects is one based on the human hand. In the more near future, a complex hand mechanism may be used as part of a robotic system to complete tasks that would normally be completed by a human but the task is more difficult or more dangerous to complete due to the environment. Specifically this is being researched for applications in space (Farry et al, 1996, and Carrozza et al, 2002). These projects investigate the possibility of allowing astronauts to complete certain tasks that would normally be extra-vehicular activities by remotely operating a hand mechanism mounted on a robotics arm system. This would eliminate the difficulty associated with trying to manipulate objects while wearing a pressurized glove and the elevated level of danger of an extra-vehicular activity. Although more of a side effect than a direct application, another field that could benefit from development of this hand mechanism is digital human modeling. Digital human modeling tries to replicate what a human does in a virtual environment for the purpose of studying how a human might interact with products in order to further refine these products before they are built. Development of a control system that is sufficiently

74 65 advanced enough to replicate human motion when grasping an object, and do so with a minimum of effort from the user, would necessarily require an intense study of all the processes that happen when a human grasps and manipulates an object. Processes that require better understanding include how the human brain coordinates individual finger movements and inter-coordination of the fingers to work together without conscious thought, how tactile feedback plays into this, and how humans unconsciously decided on a grasping strategy based on the object s shape, size, and weight, and the task that is to be accomplished. A better understanding of these things would be important in developing a control system for the hand. This would benefit the field of digital human modeling, which also aims to mimic what humans do in real life, but tries to replicate it in a virtual environment. 5.2 Results A fifteen DOF hand has been developed. It has all the major DOF s in the flexion / extension plane that a human hand has, which is many more than any commercially available model and competitive with other complex hand mechanisms currently being researched. A system to actuate it has been developed that couples the joints together in sets of thee, reducing the number of motors necessary and therefore simplifying the required electronics. Motion control and a hierarchical control scheme have also been developed. Although the hand does not mimic the human hand perfectly, it is capable of performing a cylindrical grasp, a spherical grasp, and a pinch grasp (Figure 5.1 through Figure 5.5). In this section we show examples of the hand performing these grasps.

75 66 Figure 5.1 Cylindrical grasp without glove Figure 5.2 Spherical grasp with glove, one

76 67 Figure 5.3 Spherical grasp with glove, two Figure 5.4 Pinch grasp with glove, one

77 68 Finger 5.5 Pinch grasp with glove, two The grasps developed look natural can hold objects securely. The cylinder shown in Figure 5.1 is a steel bar that weighs approximately 350 grams and is 19 mm in diameter. The bar has two different surface finishes, a smooth one and a rough one. When only the smooth surface is gripped, the bar will slip if it is held in the vertical position. The bar stops slipping as soon as the rougher surface slides in between the index finger and thumb. However this test was conducted with out the use of the cosmetic glove, which has a much higher coefficient of friction than the aluminum surface of the hand. The device weighs approximately 388 grams. This includes the weight of the fingers and hand body and does not include the glove, actuation system, or electronics. The aluminum hand body consists of a large percentage of the weight of the hand,