CROSS-REFERENCED APPLICATIONS

[0001] This application claims priority to U.S. Provisional Patent Application No. 62/543,823 filed on August 10, 2017, the disclosure of which is incorporated herein by reference in its entirety.

STATEMENT OF GOVERNMENT SUPPORT

[0002] This invention was made with Government support under grant number 1532239 awarded by the National Science Foundation. The government has certain rights in this invention.

FIELD OF THE INVENTION

[0003] This invention relates to a medical device that fits over a hand of a patient to assist with exercises. More particularly, this invention relates to an exoskeleton that can be fitted over a hand and the control of joints in the exoskeleton for assisting with movement of the hand.

BACKGROUND OF THE INVENTION

[0004] Strokes are a leading cause of disability in adults. Approximately 795,000 people experience a new or recurrent stroke annually in the United States. In the aftermath of a stroke, up to 80% of victims have some form of hemiparesis that involves weakened movement of the contralateral limbs. Another possible result of a stroke is spasticity that is the impairment of communication between the brain and muscles that results in tight muscle cramps or spasms causing a reduction in coordination and muscle movement. Furthermore, studies have shown that about 28% of hemiparetic patients were spastic three months after having a stroke. Another study found that spasticity had a prevalence in 38% of patients studied twelve months after a stroke. This impaired movement and dexterity in an upper limb and/or hand greatly affects the ability of a patient to perform normal day-to-day activities. However, it is believed that the brain can reorganize and adapt to the damage caused by a stroke with the proper training. Therefore, patients of a stroke may be given physical therapy following a stroke that includes exercises that emphasize repeated movements in order to allow the brain to relearn the motor functions lost due to the stroke.

[0005] To aid in this physical therapy, a hand exoskeleton may be used. A hand exoskeleton is an assistive device worn by a stroke patient to assist in therapeutic exercises during rehabilitation. The device aids in, guides, and/or reacts in specific ways to motions made by a patient in order to facilitate physical therapy exercises in order to allow the patient to relearn the lost motor function in a hemiparetic limb. These devices can also be used to capture quantitative data measurements to help characterize and guide a patient's recovery.

SUMMARY OF INVENTION

[0006] This application is directed hand exoskeletons and the methods of use as well as methods of adjusting the size to accommodate a wide range of users for a variety of purposes.

[0007] Many embodiments include a hand exoskeleton manipulator that has a base and a plurality of finger 3R planar mechanisms extending out of the base. The finger 3R planar mechanisms have an interface that connects to a finger, where each of the plurality of 3R planar mechanisms is positioned on a dorsal side of a finger of a user. Each of the 3R planar mechanisms has a plurality of linkages aligned to form a longitudinal axis and a plurality of joints. At least one internal joint exists connecting adjacent ends of two of the plurality of linkages and is rotatable about an axis substantially perpendicular to the longitudinal axis of the plurality of linkages. There is also an end joint connecting an end of a first of the plurality linkages to the base which is rotatable about an axis substantially perpendicular to the longitudinal axis of the plurality of linkages Furthermore, a first interface is present that connects to a last of the plurality of linkages distal from the base and connects to a finger. Each mechanism has a thumb 3R planar mechanism extending out of a side the base which has an interface that connects to a thumb of the user where the thumb joint is positioned over a dorsal side of the thumb. Each thumb mechanism also has a plurality of thumb linkages to form a thumb longitudinal axis and a plurality of thumb joints including. Each of the joints has at least one internal thumb joint connecting adjacent ends of two of the plurality of linkages and is rotatable about an axis substantially perpendicular to the thumb longitudinal axis of the plurality of thumb linkages. Each mechanism further has an end thumb joint that connects an end of a first of the plurality of thumb linkages to the base and is rotatable about an axis substantially perpendicular to the thumb longitudinal axis of the plurality of linkages. There also exists an origin affixing the end thumb joint to the base and that is rotatable about an axis substantially perpendicular to the axis of rotation of the end joint and an interface that connects to a last of the plurality of thumb linkages distal from the base and connects to the thumb of the user.

[0008] In order to actuate the hand exoskeleton described above many embodiments include a plurality of actuators for actuating joints where each of the plurality of actuators causes one of the joints in one of the plurality of finger 3R planar mechanisms or one of the joints in the thumb 3R planar mechanisms to rotate and wherein two of the joints in each of the plurality of finger 3R planar mechanisms and the thumb 3R planar mechanism are actuated by the plurality of actuators.

[0009] Many embodiments also include a control system that controls the plurality of actuators to cause rotations in at least one joint of the plurality of finger 3R planar mechanisms and the thumb 3R planar mechanism to move a hand of the user.

[0010] In other embodiments, the end joint is slidably connected to the base such that the plurality of connected linkages may slide along the axis of rotation of the end joint.

[0011] In still other embodiments, the first interface has a second companion interface slidingly connected to the first interface and configured to connect to an adjoining finger, such that as the manipulator moves the finger connected to the first interface the adjoining finger is allowed to slide freely alongside the finger connected to the first interface, thereby increasing comfort in motion.

[0012] In yet other embodiments, the first interface has a third companion interface.

[0013] In yet still other embodiments, three of the joints in each of the plurality of finger 3R planar mechanisms and the thumb 3R planar mechanism are actuated by the plurality of actuators.

[0014] In other embodiments, the hand exoskeleton manipulator has a plurality of load sensors wherein each of the plurality of load sensors are configured to create load data and wherein each of the plurality of load sensors are disposed on at least one of the plurality of linkages and in communication with the control system wherein the control system is configured to receive and utilize the load data to adjust the overall force distributed by the plurality of actuators.

[0015] In still other embodiments, the plurality of load sensors are the linkages.

[0016] In yet other embodiments, the hand exoskeleton manipulator has a plurality of encoders wherein each of the encoders are disposed in the plurality of joints and configured to create location data and in electrical communication with the control system wherein the control system is configured to receive utilize the location data such to adjust the overall control of the position of the plurality of linkages.

[0017] In still yet other embodiments, the actuators utilize Bowden cables.

[0018] In other embodiments, the plurality of actuators are disposed within an actuator support mechanism and wherein the actuator support mechanism further comprises a portion for supporting a plurality of actuator motors each of the plurality of actuator motors corresponds to a plurality of actuator capstans disposed in a separate portion of the actuator support mechanism and configured to be in mechanical communication with the base and the plurality of linkages.

[0019] In still other embodiments, at least a portion of the actuator capstans are larger than another portion of the plurality of actuator capstans.

[0020] In yet other embodiments, the hand exoskeleton has a rotational joint disposed on the last of the plurality of linkages distal from the base and configured to rotate about an axis that is substantially perpendicular to the axis of the last of the plurality of linkages distal from the base.

[0021] Many other embodiments include a method for treating a patient where the patient's therapeutic needs are determined. Once the needs are determined the type and size of orthopedic hand exoskeleton is determined. A hand exoskeleton in accordance with the various embodiments described herein may then be selected and utilized in accordance with the respective therapeutic needs of the patient. The hand exoskeleton according to the embodiments represented herein may then be fitted to the patient's hand for the desired therapeutic treatment. Furthermore, many embodiments may include an extended therapy program where the exoskeleton may be used until the desired outcome is reached. [0022] Other embodiments include a method for producing a hand exoskeleton where the length of the smallest finger that will be used by the hand exoskeleton is determined such that the finger has at least 3 joints and segments that will correspond to various linkages in the hand exoskeleton. Many embodiments include a determination of the desired workspace for the smallest finger. Once the workspace of the smallest finger is determined a plurality of sets of link lengths comprising a first link, a second link, and a third link, will then be determined, wherein the plurality of sets of link lengths are determined based on the determined length of the smallest finger. Each set of link lengths is assigned a design score based on link length and isotropic functionality of the link lengths and movement within the desired workspace. Finally a hand exoskeleton may be produced having a design score which corresponds to the desired length of the links and the amount of movement within the desired workspace in accordance with the various embodiments described herein.

[0023] Additional embodiments and features are set forth in part in the description that follows, and in part will become apparent to those skilled in the art upon examination of the specification or may be learned by the practice of the disclosure. A further understanding of the nature and advantages of the present disclosure may be realized by reference to the remaining portions of the specification and the drawings, which forms a part of this disclosure.

BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The description will be more fully understood with reference to the following figures, which are presented as exemplary embodiments of the invention and should not be construed as a complete recitation of the scope of the invention, wherein:

[0025] Fig. 1 illustrates an example of a HEXORR hand exoskeleton.

[0026] Fig. 2 illustrates a Hand Exoskeleton (HX) device that is an example of a hand exoskeleton with a segmented interface.

[0027] Fig. 3 illustrates a BioHED device that is an example of a passive hand exoskeleton.

[0028] Fig. 4 illustrates an Actuated Finger Exoskeleton (AFX) device that is an example of a hand exoskeleton that uses cables and pulleys.

[0029] Fig. 5 illustrates an example of an exoskeleton that uses Bowden cables [0030] Fig. 6 illustrates the IntelliARM device that is an example of an exoskeleton having a motor driving a linkage is a four- bar linkage.

[0031] Fig. 7 illustrates an example of an exoskeleton having a number of soft robotic systems involving soft actuation.

[0032] Fig. 8 illustrates a hand exoskeleton manipulator in accordance with an embodiment of the invention.

[0033] Fig. 9 illustrates a kinematic diagram of a 3R planar mechanism controlling a single finger in accordance with an embodiment of the invention.

[0034] Fig. 10 illustrates a kinematic schematic of the degrees of freedom of finger manipulators in accordance with an embodiment of the invention.

[0035] Figs. 1 1 a-1 1 d illustrate configurations of finger interfaces for a hand exoskeleton manipulator in accordance with an embodiment of the invention.

[0036] Fig. 12 illustrates sliding link origins for the finger 3R planar mechanisms in accordance with an embodiment of the invention.

[0037] Fig. 13 illustrates an axis of rotation for a rotatable 3R planar thumb mechanism in accordance with an embodiment of the invention.

[0038] Fig. 14 illustrates an actuator pack in accordance with an embodiment of the invention.

[0039] Fig. 15 illustrates link sensors in accordance with an embodiment of the invention.

[0040] Fig. 16 illustrates a plot of the design scores for hand exoskeleton manipulators having different link lengths in accordance with various embodiments of the invention.

[0041] Fig. 17 illustrates a plot of designs score for hand exoskeleton manipulators having a set L2 linkage in accordance with various embodiments of the invention.

[0042] Fig. 18 illustrates a hand exoskeleton with linkages for a 3R planar mechanism that are experimentally determined in accordance with an embodiment of the invention.

[0043] Fig. 19 illustrates plots of design scores generated using the modified design score equation that result from increasing the weight factor B in accordance with an embodiment of the invention.

[0044] Fig. 20 illustrates joint angles along with the end effector force of a 3R planar mechanism in accordance with an embodiment of the invention. [0045] Fig. 21 illustrates planes of the linkages as well as an end effector force vector measured by the sensor on link 2 in a hand exoskeleton manipulator in accordance with an embodiment of the invention.

[0046] Fig. 22 illustrates an end effector force vector along with the force component measured by a sensor on Link 2 of a hand exoskeleton manipulator in accordance with an embodiment of the invention.

[0047] Fig. 23 illustrates a block diagram of components of a hand exoskeleton manipulator and control system used to control the hand exoskeleton manipulator in accordance with an embodiment of the invention.

DETAILED DESCRIPTION

[0048] Turning now to the drawings, an actively actuated three-fingered dexterous hand exoskeleton for use in stroke patient rehabilitation in accordance with some embodiments of the invention is described. A hand exoskeleton in accordance with many embodiments includes a mechanical device, an actuation system and a control system. In accordance with a number of embodiments, a kinematics-based optimization method may be used to design the manipulators of the exoskeleton. A more complete description of a hand exoskeleton including various aspects of the design including manipulator kinematics, interfaces with the patient, and an actuation system as well as various preferable functionalities are described in detail below.

[0049] Hand exoskeletons have taken on many forms and use different mechanisms to help the user perform different types of motions and grasps. The human hand has four fingers, each with four degrees of freedom including of the rotations of: Metacarpophalangeal (MCP), Proximal Interphalangeal (PIP), and Distal Interphalangeal (DIP) joints in addition to adduction/abduction. The thumb has five degrees of freedom— three rotations of the joints, adduction/abduction, and opposition/apposition. Hand exoskeletons are typically designed to aid in and train various motions of the hand. One type of hand exoskeleton has been created to aid in the specific motion of the binary opening and closing of the hand. One example of this type of hand exoskeleton is a HEXORR that is shown in Fig. 1.

[0050] As shown in Fig. 1 , the HEXORR has a mitt-like interface with a hand. This is because, in general, the interface between a hand and a particular exoskeleton depends on the motion(s) that the exoskeleton supports. In the case of the HEXORR, a mitt-like interface is sufficient to assist in the opening and closing of the hand. However, exoskeletons that facilitate finer motions of the fingers including, but not limited to, flexion and extension of the fingers; and pinching typically require interfaces that are better tailored for these purposes. Interfaces for these types of motions are usually segmented. An example of a hand exoskeleton with a segmented interface is the Hand Exoskeleton (HX) shown in Fig. 2. To adjust for different hand sizes in different patients, hand exoskeletons that operate very closely to the movement of a user often incorporate features such as passive prismatic joints and adjustable link lengths in order to accommodate the variation in finger sizes among patients.

[0051] In terms of actuation, some actuator systems in hand exoskeletons leave one or several degrees of freedom of the hand passive while other degrees of freedom are actively actuated by motors. Some passive hand exoskeletons may use interchangeable or adjustable springs or cords to tune the resistances of certain training motions for a user. An example of a passive exoskeleton is the BioHED device shown in Fig. 3. The BioHED device uses an exotendon layout of springs. The overall tension of the springs is adjusted by a servo motor.

[0052] There are several styles of active actuation for hand exoskeletons. Some hand exoskeletons use a style of active actuation that implements a tendon-like system to aid in flexion and extension of the entire hand in an agonist-antagonist manner. Other types of hand exoskeletons incorporate systems of cables and pulleys. An example of a hand exoskeleton that uses cables and pulleys is the Actuated Finger Exoskeleton (AFX) device shown in Fig. 4. The AFX in Fig. 4 also includes links that are geared and can move past one another. However, this is not typical, most designs implement linkages with fixed attachment points. Some cable-driven systems use Bowden cables to remotely actuate the device. An example of an exoskeleton that uses Bowden cables is shown in Fig. 5. The advantage of a system that uses Bowden cables is that the distance between the actuator pack and the device itself can be increased. The ability to increase this distance can allow the actuator pack to be placed remotely from the patient and prevents the patient from having to bear the weight of the actuators while wearing the exoskeleton. The use of Bowden cables can also remove the need for a tensioning mechanism between the joints of the fingers, which move relative to one another.

[0053] Other options for active actuation include, but are not limited to a motor directly driving a linkage and soft robotic systems. An example of an exoskeleton having a motor driving a linkage that is a four- bar linkage of the IntelliARM shown in Fig. 6. An example of an exoskeleton having a number of soft robotic systems involving soft actuation is shown in Fig. 7. The exoskeleton in Fig. 7 includes chambers that are actuated via pressurized fluid.

[0054] The following table provides a summary of relevant past research in hand exoskeleton devices. As evidenced by the variety of devices that have been produced in the past, creating an assistive or rehabilitative exoskeleton for the hand is a complex problem with many factors that must be taken into account and addressed by the design.

Design Considerations

[0055] Hand exoskeletons, in accordance with many embodiments of the invention may comprise one or more of the following functionalities. First, many embodiments may have adaptable features to accommodate a wide range of hand sizes; preferably capable of accommodating up to the 95 ^th percentile of hand size. Second, many embodiments may allow for the user to articulate any number of grasp types and motions in accordance with user comfort and patient application. In other words, many embodiments may also be adaptable to the specific rehabilitation of the patient. Third, many embodiments may be configured such that the hand exoskeleton may be disposed or integrated onto an existing upper limb exoskeleton such as for example, the EXO-UL7. Finally, many embodiments may be configured such that the actuators may be able to control the hand exoskeleton remotely thereby reducing the overall weight of the device. General Form of a Hand Exoskeleton

[0056] In accordance with some embodiments of the invention, a hand exoskeleton manipulator is comprised of three serial Three-Rotation (3R) planar mechanisms. The use of 3R planar mechanisms allow the manipulators to have sufficient flexibility to reach the entire areas of digits' in-plane workspaces without interfering with the movements of the digits. The choice to use the three manipulators stems from 95% of human grasps being achievable with three digits, namely a thumb and two fingers, and a human hand is redundant in having five digits.

[0057] A hand exoskeleton manipulator in accordance with an embodiment of the invention is shown in Fig. 8. In Fig. 8, hand exoskeleton manipulator 800 includes a base 805. Two 3R planar mechanisms 801 and 802 extend out of the base on the dorsal side of the hand above the MCP joints of a finger. Although, two 3R planar mechanisms that extend out over the fingers are shown in this embodiment, up to four 3R planar mechanisms may extend out over the fingers in accordance with various other embodiments of the invention. A third 3R planar mechanism 810 extends out of base 805 on the dorsal side of the thumb above the CMC joint of a thumb and can be used to control the movement of the thumb. All manipulators 801 , 802, and 810 can control the digits from the dorsal sides of the digits to prevent interference at the sides of the digits and to allow the user to grasp objects while using the exoskeleton. The end effector 820 of each of the manipulators 801 , 802, and 810 interfaces with the digit tip and operates in the plane of the digit. The two finger linkages can control the fingers in different groupings. One of the finger linkages will typically be controlling the index finger. In many embodiments, one of the manipulators controls the index finger and another manipulator can control the little finger. A kinematic diagram of a 3R planar mechanism controlling a single finger in accordance with an embodiment of the invention is shown in Fig. 9. The same principal applies to the thumb mechanism. However, the CMC joint of the thumb replaces the MCP joint of a finger. Each of the revolute joints shown rotates about an axis pointing out of the page. [0058] The link lengths of each of the three mechanisms 801 , 802 and 810 in accordance the shown embodiment are recorded below in the first table. These link lengths were obtained via an optimization method detailed below. The lengths used for the digits' bones were those of the 95th percentile based on a hand anthropometry study of U.S. Army Personnel, these values are shown below in the second table. Although certain link lengths are represented in the tables below it should be understood that any link length may be utilized based on the desired application and end user of the exoskeleton.

Degrees of Freedom

[0059] In accordance with the embodiment illustrated in Fig. 8, Joints 1 and 2 of each 3R planar manipulator are actively actuated, while Joint 3 remains passive. However, any combination of the three joints may be actively actuated in accordance with various other embodiments of the invention. In the shown embodiment, joints 1 and 2 are actively actuated to allow the hand exoskeleton manipulator 800 to control the position of a digit within the workspace of the digit. Actuating only one joint would give the least amount of control over the digit position. Therefore, more than one joint can be actuated. One option is the active actuation of all three joints in accordance with many embodiments of the invention. This would provide the greatest control over the position of the end effector and the digit tip. However, user comfort typically entails controlling the third joint in such a way that the orientation of the tip of the digit is comfortable to the user. This is a difficult task because different users will have digits of varying sizes, and predicting what third joint angles and end effector orientations are comfortable to the user in different areas of the workspace can be challenging. In addition, actively actuating all three joints often requires the greatest number of motors overall for the hand, which can make both the hand exoskeleton and the actuator pack larger and heavier. Therefore, it is preferred to leave at least one joint passively actuated rather than actively actuating all three joints.

[0060] In accordance with many embodiments, two joints are actively actuated. Through optimization, which will be discussed in detail below, it was found that the third link is the shortest in length for each 3R mechanism. This implies that the angle of the third joint has the least influence on the position of the end effector and caters mostly towards the orientation of the end effector. It follows that actuating the first two joints, which correspond to the first and second links, affords the most control over the position of the end effector. Leaving the third joint passive can allow a user's digits to be naturally oriented in a way that the user finds comfortable.

[0061] Hand exoskeleton manipulator 800 may also provide adduction and abduction of the fingers. These motions are not actively actuated in the interest of reducing the device's overall size and weight in accordance with the shown embodiment. However, not accommodating adduction and abduction restricts the natural motions of the fingers and makes movement stiff and entirely in plane. As such, adduction and abduction may be passively allowed by the hand exoskeleton 800 over small angle ranges via a rotational axis on the third link of each finger mechanism, in accordance with various embodiments. A kinematic schematic of the degrees of freedom of the finger manipulators illustrating the rotational axis for abduction and adduction, in accordance with various embodiments of the invention, is shown in FIG. 10. Adduction and abduction motions are not left entirely passive over sizeable angle ranges, as doing so typically involves longer third links to create space for the motions or additional bearings at the linkage bases. Both of these additions result in a larger and heavier device. Adduction and abduction for the thumb 3R planar mechanism in accordance with the embodiment shown in Fig. 8 are neither actively controlled nor passive to any extent. However, the orientation of the base of the hand exoskeleton manipulator 800 may be adjustable allowing a user to preset the degree of adduction and abduction prior to operation according to various embodiments.

[0062] Various embodiments of a hand exoskeleton are described above. However, alternative embodiments of hand exoskeletons may include different number of joints, different number of active and/or passive joints, and different number of joints with multiple degrees of freedom depending on the preferred embodiments of a particular exoskeleton.

Grouping of Fingers

[0063] A number of features of the exoskeleton 800 allow the wearer to train different types of grasps, as opposed to only supporting a binary opening and closing of the hand. One such feature in accordance with the shown embodiment of the invention is the use of interchangeable end effector interfaces for the two finger 3R planar mechanisms. In accordance with many of these embodiments, the finger interface is the portion of the third link that is strapped to the fingers of a user. These finger interfaces can be designed in multiple configurations in order to accommodate motions that require the four fingers to be grouped in specific ways. For example, a pinch typically requires one 3R planar mechanism to control the index finger and the other 3R planar mechanism to control the grouped middle, ring, and little fingers along with the thumb 3R planar mechanism. A thumb-2-finger grasp typically requires one 3R planar mechanism to control the grouped index and middle fingers and the other 3R planar mechanism to control the grouped ring and little fingers. The four configurations of finger interfaces that may be utilized in conjunction with the exoskeleton in accordance with the embodiment illustrated in Fig. 8 are shown in Figs. 1 1 a-1 1 d.

[0064] As seen in Fig. 1 1 a and 1 1 d, various embodiments of a 2-Finger Slider interface are illustrated. The two finger slider interface allows the two adjoining fingers to slide relative to each other during flexion and extension. Such motion is more natural to how finger joints move thus allowing for greater comfort for the end user. Fig. 1 1 b illustrates a 3-Finger Slider interface that would function in a similar manner as the two finger slider in Figs. 1 1 a and 1 1 d. Such embodiment would allow the ring and little fingers to slide relative to one another for example. Due to the large difference in length between fingers, allowing for sliding motion to occur as illustrated in the various embodiments is preferable for human interaction to increase comfort. The finger slider mechanisms described herein may be utilized with one or more finger planar mechanisms described herein.

[0065] Different finger configurations and interfaces in accordance with various embodiments of the invention are described above. However, other finger configurations and/or types of interfaces in accordance with various other embodiments of the invention to satisfy the preferred embodiments of particular hand exoskeletons.

Sliding Finger Mechanism Bases

[0066] A feature incorporated into hand exoskeletons in accordance with some embodiments of the invention that enables multiple grasps is sliding link origins 1210 for the finger 3R planar mechanisms as shown in Fig. 12. The sliding link origin 1210 can work in conjunction with different finger groupings as described in the above description. Since the end effectors of the mechanisms are controlled in a plane, changing which fingers each of the two 3R planar mechanisms controls involves linkage origins 1210 that move laterally above the knuckles in order to avoid patient discomfort. For example, if one 3R planar mechanism controls just the index finger, the linkage origin can be slid such that it is directly above the MCP joint of the index finger. The origin of the other mechanism, which controls the grouped middle, ring, and little fingers, can be positioned around the MCP joint of the ring finger. Ruler markings on the side of a track on which the link origin slides can allow the users of the exoskeleton to measure and record the positions of the link origins for each mechanism in accordance with a number of embodiments.

Rotatable Thumb Mechanism

[0067] Another feature in accordance with various embodiments of the invention can support different grasps is a rotatable thumb mechanism origin 1310. The origin of the thumb linkage can be loosened and rotated to a certain desired angle in increments of 22.5°. The origin 1310 according to various embodiments may also be fixed at the desired angle before the user puts on the exoskeleton. The axis of rotation is depicted in Fig. 13. The rotatable origin of the thumb mechanism can maintain a desired trajectory for the thumb during exercises. For example, a thumb of a stroke patient experiencing spasticity tends to default to a key grasp position. Therefore, fixing the thumb 3R planar mechanism at different degrees of adduction and abduction can allow the patient to practice other types of grasps.

[0068] The above features of hand exoskeleton manipulators in accordance with some embodiments of the invention allow a hand exoskeleton to assist a patient in performing various power and precision grasps, pinches, etc. Based on the taxonomy of grasps, the user can perform the following grasps using a hand exoskeleton manipulator in accordance with some embodiments of the invention: power grasps including, but not limited to a lateral inch; six prismatic grasps including: large diameter, small diameter, medium wrap, adducted thumb, and light tool; and precision grasps including, but not limited to, thumb-4 finger, thumb-2 finger, and thumb-index finger. Although various embodiments of a hand exoskeleton are described above with reference to Figs. 8-12 are described above, other embodiments that combine, omit, and add additional features and/or support any variety hand motions are possible.

Accommodation of Various Hand Sizes

[0069] Due to the variation in human hand sizes and the positioning of different anatomical features on the hands, hand exoskeleton manipulators in accordance with some embodiments of the invention are able to accommodate as many different users as possible. To do so, hand exoskeleton manipulators in accordance with several embodiments of the invention are configured such that each linkage accommodates a 95th percentile hand. For purposes of this discussion, hand measurements were taken from an anthropometry study involving United States Army personnel, T. Greiner, "Hand Anthropometry of U.S. Army Personnel," 1991 , and the 95th percentile values for the male subjects of the study. The most important dimensions for the hand exoskeleton manipulator are (i) the lengths, diameters, and breadths of the proximal, medial, and distal phalanges for each of the fingers, (ii) the lengths, diameters, and breadths of first metacarpal, proximal phalanx, and distal phalanx for the thumb, and (iii) the overall length and breadth of the hand. In accordance with some embodiments, a hand exoskeleton may be configured to the size and shape of a specific hand and may be based upon a 3D print of the hand. In accordance with many embodiments, hand exoskeletons may be configured to handle any variety of hand sizes.

[0070] The sliding mechanism origins and finger interfaces of the hand various exoskeleton manipulator described above can also contribute to accommodating hands of various sizes. The sliding manipulator bases can allow the mechanisms to be positioned above the appropriate knuckles. These positions vary depending on the size of the hand of an individual user. The sliding finger interfaces can allow hands to fit the interfaces regardless of the variability in the difference in length of the ring and little fingers.

[0071] Another advantage of using interchangeable finger interfaces in accordance with some embodiments of the invention is that the interfaces are made of ABS plastic and may be 3D printed in accordance with a number of embodiments. This can allow for the quick creation of interfaces of various sizes, thicknesses, and curvatures. In this way, an interface can be generated that fits patients with particularly thin or thick fingers comfortably.

[0072] In order to increase the likelihood that the 3R planar mechanisms can reach any point within the workspace of a 95th percentile finger, appropriate link lengths can be chosen for the exoskeleton in accordance with many embodiments of the invention. In order to find the desired balance between workspace coverage, link stiffness, and mechanism length, an optimization process can be performed to determine the link lengths for each mechanism. Various optimization processes are described below and a case study of the index finger manipulator is presented.

Actuation

[0073] A hand exoskeleton manipulator in accordance with some embodiments of the invention may be actuated by a motor pack separate from the manipulator. The actuator pack may be mounted on an existing component of an existing EXO-UL7 upper limb exoskeleton in accordance with many embodiments of the invention. Power can be transmitted from the actuator pack 1400 to the actuators in joints of the 3R planar mechanisms through a system of capstans and Bowden cables. A rendering of the uncovered actuator pack 1400 in accordance with an embodiment of the invention is shown in Fig. 14. During operation, this pack may be covered with a 3D-printed ABS shell for protection or any other type of cover. It may also be left open and uncovered. There are two main advantages of providing power in this manner. The first main advantage is that it allows for flexible and unpredictable routing of the cables. Throughout operation, the hand exoskeleton manipulator opens and closes. As such, the entire hand of the user can move relative to the actuator pack 1400. The flexible sheathing system is capable of accommodating curved paths in 3D space that match this movement. The second main advantage is that the power delivery system can maintain constant tension despite the user moving the hand exoskeleton manipulator. With fixed attachment points and a fixed sheath length, the sheath bends and adjusts the path between the actuator pack and the manipulator joints such that the distance the cable travels is constant (provided that the path is no longer than the sheathing itself).

[0074] A disadvantage of the Bowden cable actuation system is that the system is subject to friction between the sliding cable and the inner material of the sleeve. This friction can be complex and can vary with bend angle and overall cable tension. Furthermore, Bowden cable systems can also be susceptible to backlash as the cable slides within the sheath.

[0075] To address these limitations, the actuator pack motors 1410 in accordance with a number of embodiments of the invention may be oversized so that frictional losses do not result in the manipulator joints receiving insufficient torque from the motors. In addition, the difference between the cable diameter and the inner diameter of the sheathing may be minimized in accordance with several embodiments.

[0076] In the illustrated embodiment, the cable chosen is stainless steel of 7x49 construction with a breaking strength of 170 lbs. The 7x49 construction can allow for flexibility. The high breaking strength can provide three advantages. Firstly, the high breaking strength gives the power transmission system a large safety factor relative to the grasp strength of a human hand to prevent failure. This safety factor can also account for the cable adhering to the curve around a capstan typically does not achieve 100% of its breaking strength while straight. Secondly, the durability of the Bowden cable can help alleviate the need to change cables frequently. Thirdly, this cable and its high breaking strength have a correspondingly large diameter (0.044 in the illustrated embodiment), which fits into the inner diameter of the conduit (inner diameter of 0.072) such that smooth motion is maintained while minimizing backlash.

[0077] The actuators, in accordance with some embodiments may be Maxon RE 13 motor combinations with a maximum output torque of 2.0 Nm. From Fig. 14, it can be seen that capstans attached to the upper row 1420 of motors are larger, with 14mm diameters, than the capstans on the bottom rows 1430, with 9mm diameters. The difference in the respective diameters of the capstans can allow the motors to be in a compact, stacked configuration without the cables from the top capstans 1420 interfering with the cable from the capstans in the bottom row 1430. However, the joints actuated by the bottom row of motors in this configuration may receive only a fraction of the motors' output torques. In the illustrated embodiment, the aforementioned joints have a maximum output torque of 1 .28 Nm. Therefore, these motors may control Joint 2 of the manipulators.

[0078] Actuators in accordance with some embodiments of the invention are described above. However, other actuator systems and/or configurations can be used in accordance with some other embodiments of the invention.

Sensors

[0079] In accordance with many embodiments, the force exerted by the user's finger(s) on the end effector is measured and provided as the input to a control system as discussed below. To sense the force, each of the three mechanisms may be configured with a bending beam load sensor, such as the sensor provided by FUTEK Advanced Sensor Technology, Inc., incorporated into one of the links of the mechanism in accordance with many embodiments. For the finger mechanisms, the force sensor can be part of the second link in accordance with a number of these embodiments. The second link is chosen based on the proximity of the link to the end effector and to simplify the calculating of the end effector force based on the joint angles read by the system encoders. For the thumb mechanism, the bending beam load sensor can be incorporated into the first link of the manipulator in accordance with many embodiments. In order to aid in the overall compactness of the exoskeleton, it is preferable to incorporate the bending beam load sensor into the longest linkage. Through experimenting with physical prototypes, it was found that making the sensor part of the second link prevented the thumb linkage from being made as compact as possible. Therefore, the bending beam load sensor can be made part of the first linkage, which typically has the longest overall length. The implementation of these sensors into the links in accordance with an embodiment of the invention is shown in Fig. 15. In Fig. 15, sensor link 1501 is for the finger mechanism and sensor link 1502 is for a thumb mechanism.

[0080] In accordance with some of these embodiments, the active joints of the mechanisms are sensed by encoders attached to the joint motors and gearboxes (Maxon). The passive joint of the manipulator can be sensed using a magnetic incremental encoder (Renishaw, Inc.).

[0081] Various sensors in accordance with some embodiments of the invention are described above. However, other types of sensors and/or sensor configurations may be used in accordance with various other embodiments.

Optimization Process

[0082] In accordance with some embodiments of the invention, an optimization method is performed to determine the link lengths of the two finger mechanisms and the thumb mechanism of a hand exoskeleton manipulator. The goal of various embodiments of the invention is for the device to accommodate a variety of hand sizes with fixed link lengths and achieve the best kinematic performance within a desired workspace. An optimization process in accordance with many embodiments of the invention is focused on both kinematics and workspace coverage. In contrast, previous work in optimization for hand exoskeletons has focused on maximizing torque transmission on the fingertip being manipulated.

[0083] The overall objective of an optimization process in accordance with some embodiments is to determine the set of link lengths for the 3R planar mechanisms that will give the best kinematic performance within the workspace of the finger. The methodology behind the optimization is as follows. The algorithm receives as inputs the variables for the potential link lengths L ₁ , L2, and L3. The algorithm tests all of the different combinations of link lengths and calculates a design score for each combination that is based on parameters that will be discussed in below. Via this brute force method, a theoretically optimal design is determined. Following the calculations, some criteria that automatically invalidate a design if they are not met for the combinations are checked.

[0084] The first criterion is overlapping workspace area. The workspace of the 3R planar mechanism must completely overlap the workspace of the finger, i.e., the end effector of the mechanism must be able to reach anywhere the fingertip could move. To address this, the algorithm calculates the boundary and size of both workspaces to ensure that they overlap completely. The second criterion is inverse kinematics. If the inverse kinematics solutions for a mechanism do not exist for all points in the finger's workspace, the design is invalidated. The third criterion is interference between the mechanism links and the finger. If the inverse kinematics solution for the set of potential link lengths shows that in order to reach the fingertip, a joint of the mechanism must lie within the flesh of the finger, the design is marked as unacceptable, as this indicates there would be interference between the user and the mechanism links.

[0085] First, the algorithm calculates the workspace of the smallest finger that will be manipulated by the linkage, as this finger is the one that limits the motion of the other fingers, if any, that are also attached to the end effector. For a finger 3R planar mechanism in accordance with some embodiments, the smallest finger is often the index finger, and for the other mechanism, the smallest finger is the little finger. For a potential set of link lengths (L ₁ , L2, L3), the mechanism isotropy is determined for the joint spaces that position the end effector within the workspace boundary of the finger. The information obtained from this process can be used to calculate a design score for the potential set of link lengths. This process is then repeated for each of the combinations of links within a certain size range for each link in the mechanism. Evaluating this data for both finger mechanisms and the thumb mechanism can yield theoretically optimal link lengths for the exoskeleton. Determining Design Score

[0086] One component of a design score in accordance with some embodiments is kinematic performance. The analysis of kinematic performance in the optimization process in accordance with many embodiments uses a mechanism isotropy (ISO) performance measure as one parameter in the design score, which grades each potential mechanism. The mechanism isotropy performance measure is a measure of the ability of a mechanism to move an end effector in any direction. The mechanism ISO performance measure is a function of the joint angles (θ ₁ , θ2, Θ3) and ranges from 0 to 1 . Mechanism isotropy may be defined by the following equation:

[0087] Where λ _min and _max are the minimum and maximum eigenvalues of the Jacobian matrix. An isotropy value of zero typically indicates a singularity and the loss of a degree of freedom; while an isotropy of 1 often indicates that the end effector can move equally well in all directions.

[0088] Potential mechanisms should have an increase in design score in accordance with higher mechanism isotropy values within the finger workspace. One way to incorporate isotropy into the score is by summing isotropy values over the finger workspace. Let K be defined as the set containing the discrete points within the finger workspace that have been calculated. Each calculated point has an associated isotropy value. In finding the sum of the mechanism isotropy within the workspace, an adjustment can be made to account for varying densities of calculation points in K. To address this, the finger workspace area can be discretized into a grid of cells along the x- and y- directions. The average mechanism isotropy can be calculated within each of these cells and multiplied by the area of the cell. The sum total of all the cells' area-weighted values can be taken as the sum of isotropy. A potential mechanism can be rewarded for having a higher sum of isotropy. [0089] If there is a position where the isotropy is zero within the finger workspace, the design score should be zero as well because this indicates a singularity. A potential mechanism should also have a higher design score for having a higher minimum mechanism isotropy value within the workspace compared to other sets of link lengths. These preferred embodiments are reflected in the numerator of the equation for design score.

[0090] Another factor to consider in a design score utilized in the design of a hand exoskeleton in accordance with various embodiments of the invention is that although larger links tend to lead to better system isotropy values within the workspace, the larger links can also result in the mechanism having greater mass and lower mechanism stiffness, which are unfavorable. Additionally, long link lengths increase the size of the envelope of the mechanism above the wearer's hand. This increases the potential for interference occurring between the linkages and surrounding objects or other parts of the exoskeleton during therapeutic exercises. Therefore, potential designs should be penalized for being too large. Beam theory shows that mechanism stiffness is inversely proportional to the cube of the sum of the link lengths. The final design score equation includes this in the following expression:

Modeling a Mechanism

[0091] In accordance with some embodiments of the invention, each finger and thumb mechanism of a hand exoskeleton's manipulator is a 3R planar mechanism. For a general 3R planar mechanism, the only nonzero Denavit-Hartenberg parameters correspond to all three of the link lengths and their joint angles as shown in the below table. Table 3.1 shows a complete listing of the DH parameters.

[0092] The position of the end effector is described by two coordinates (x,y) and an orientation (φ) and is given as the following forward kinematics equation:

Where

[0093] The forward kinematics yield a Jacobian matrix with respect to the ground frame as

[0094] However, using this 3x3 Jacobian matrix where every entry in the third row is 1 results in a matrix that can give negative minimum eigenvalues and positive maximum eigenvalues. The ratio of these yields negative numbers, which do not fit into an isotropy measure, which is bounded by 0 and 1 . If the two rows corresponding to the x- and y- coordinates are used as a 2x3 matrix, the singular values of the matrix J ^T J may be used. Taking q to be the position vector of the end effector and Θ as the column vector of joint angles gives:

From a singular value decomposition of the Jacobian as this becomes

U and V are unit matrices, giving:

From the above, it is apparent that J ^T J and are similar, and the eigenvalues of J ^T J are the squares of the singular values of the original Jacobian matrix J. Therefore, these values are used to calculate mechanism isotropy. [0095] The inverse kinematics for the 3R planar mechanism are shown below. These equations can be used in an optimization to check that the joints of a potential design do not interfere with the flesh of the finger when the finger is in the clenched position.

Modeling of the Finger

[0096] The index finger can also be modeled as a 3R planar mechanism in accordance with some embodiments of the invention. As such, the same kinematics equations apply. Each relevant joint and bone in the anatomy corresponds to a joint and a link in the 3R planar mechanism model. For the fingers, the mechanism starts from the MCP joint (Joint 1 ) continues along the proximal phalanx (Link 1 ) to the PIP joint (Joint 2) further continuing along with medial phalanx (Link 2) to the DIP joint (Joint 3), and finally extending along the distal phalanx (Link 3) to the fingertip. For purposes of performing an optimization process in accordance with some embodiments of the invention, the lengths of the phalanges are set to be those in the 95th percentile for each bone. The proximal, medial, and distal phalanges may be taken to be L ₁ = 2.79 inches, L2= 1 .06 inches, and L3= 1.27 inches. The joint ranges used for each potential mechanism in the optimization calculation are given in the following table:

[0097] The hand exoskeleton physically may interface with the dorsal side of the wearer's hand via a modified glove in accordance with many embodiments. The glove material, form factor of the glove, and the physical interface between the fingertip and end effector of a mechanism restrict the finger from reaching the entire uninhibited workspace of the finger. To account for these restrictions of movement, multiple angle ranges can be defined for each of Joints 1 through 3. These angle ranges are used during the optimization in determining the finger workspace. Each joint has a primary range, a restricted range, and a secondary range. The primary range can be used to describe most of its positions. The restricted range is the range of the next joint is restricted to its secondary range. These angle ranges were obtained through observations while wearing a glove of a hand exoskeleton manipulator in accordance with an embodiment of the invention and the information is rovided in the following table.

Results of Optimization Process

[0098] A plot of the design scores of various hand exoskeleton manipulators in accordance with a number of embodiments of the invention for the range of interest of L ₁ are shown in FIG. 16. These various points in

the plot are the design scores for all possible combinations of link lengths within the given range. Due to a sensor to be included in the second link of a mechanism in the actualized hand exoskeleton, the constraint that L2 = 3.5 inches. Therefore, the relevant subset of data is the plane where L2 = 3.5 inches. The design scores on this plane are shown in Fig. 17. In Fig. 17, the white data points on this contour plot mark various link length combinations within this plane in, accordance with various embodiments, based on the criteria of workspace overlap, inverse kinematics, and joint interference as described above.

[0099] As further seen in Fig. 17, a concentration of design scores lies towards the bottom right corner of the contour plot where L ₁ is at the highest within the data range and L3 is at its lowest. Of the acceptable designs, the one with the highest design score, i.e. the one closest to this area of interest, is the one that the optimization algorithm has determined to be the optimal one. The preferable set of link lengths for some embodiments is L ₁ = 4.0 inches, L2 = 3.5 inches, L3 = 1 .95 inches. A mechanism linkage for the index finger having these dimensions was realized as a physical prototype as shown in Fig. 18. This rapid prototype was created with adjustable links to size the entire linkage as appropriate to a particular hand. The objective of this is to experimentally adjust and shape the links to make the links as short as possible while maintaining smooth movement, ensuring total workspace coverage, avoiding singularities within the workspace, and interference between the links and the index finger. The results in the set of experimentally optimized link lengths of L ₁ = 3.0 inches, L2 = 3.5 inches, and L3 = 1 .3 inches.

[00100] A comparison between a theoretically optimal set of linkages as determined by the optimization algorithm and an optimal set of linkages determined experimentally are shown in the following table. The optimization algorithm results are marked as Theoretical Unweighted. The values in parentheses are the percent differences between the actual and theoretical values of various quantities of interest.

[00101] As shown in the above table, the experimental values for L ₁ and L3 are lower than those produced by the optimization by 25.0% and 10.3% respectively. The sum of isotropy is lower by 21 .8%, and the minimum isotropy value is greater by 2.1 %. A potentially confusing result is that the design score is greater than the one predicted to be optimal by 19.9%. Two main factors contribute to these differences between the theoretical and actual values for the link lengths. The first factor is the fact that the optimization algorithm and the experimental optimization were performed for different hand sizes. As mentioned above, the optimization algorithm used a model finger comprised of proximal, medial, and distal phalanges that were all in 95th percentile in length in order to ensure that the manipulator would be able to accommodate as many individuals as possible. However, a finger with such phalanges would have an overall length that lies above the 99th percentile. In contrast, the experimental adjustment of the links was done in accordance with the fingers of an actual user having fingers that fall within the 95th percentile for overall length as opposed to the 95th percentile in length for each individual phalanx. With this in mind, the conceptual expectation is that the link lengths obtained experimentally are smaller than the ones obtained through the optimization. Therefore, it is expected that the theoretically optimized manipulator is larger than the manipulator created via a physical prototype as shown in the table. Because the design score is inversely proportional to the cube of the sum of the link lengths, this fact results in a design score that is higher than that of the theoretical set of link lengths. The reason that design obtained experimentally was not predicted by the algorithm as the optimal one is that this shorter set of link lengths was marked as unacceptable due to the optimization algorithm, as these shorter links do not have solvable inverse kinematics without any joint interference for the larger finger used in the optimization.

[00102] The second factor contributing to the differences between the theoretical and experimental values in the table is the difference in priorities between the algorithm and the experimental optimization.

[00103] The original design score equation weighs each of the terms related to isotropy equally, and the sum of the cube of the link lengths three times as much. However, when the links were experimentally adjusted, the highest priority was making the link lengths as small as possible, provided that the manipulator's movement still felt uninhibited within the finger workspace. This different priority introduces a more subjective aspect to the design process that the original design equation does not consider. The particular emphasis on reducing link size is not reflected in original design score equation given above. Therefore, there is a disagreement between the design priorities of the algorithm in accordance with some embodiments and the experimental process while designing a prototype.

Modified Design Score Equation

[00104] A modified design score equation that can be used to design hand exoskeletons in accordance with some other embodiments of the invention includes weight factors that allow for the flexibility of increasing or decreasing the relative importance of components in the design score equation. The modified equation is expressed as:

Where A and B are the weight factors for the isotropy terms and the link size terms, respectively.

[00105] Plots of design scores generated using the modified design score equation in accordance with an embodiment of the invention were in the plane where L2 = 3.5 inches was the result from increasing the weight factor B, which increases the importance of link size in the design score is shown in Fig. 19. In Fig. 19, the original hot spot region for design score fades while the region with lower L ₁ and L3 emerges as the new hot spot as B increases. This transition between hot spots takes place when the weight factor for link size B = 3.75. It follows that the acceptable design with the highest design score according to modified design score equation becomes L ₁ = 3.0 inches, L2 = 3.5 inches, and L3 = 1.3 inches. The bottom row of in the above table of results shows the comparison of percent differences in the design score equation with the weight factors. The 34.6% difference between L3 values methods can explained by physical constraints in the construction of the prototype. Specifically, in order to ensure that the proper degrees of freedom are present in the physical mechanism, in many embodiments, the third link may have two rotation axes. Due to the amount of space these axes require, L3 = 1 .75 inches is the limit to how small the third link can physically be. Overall, the approach using weight factors in accordance with some embodiments of the invention can lead to greater agreement between the results of the optimization algorithm and the experimental adjustment of the physical prototype.

[00106] As seen in Fig. 19, increasing the weight factor B beyond the point at which the two hot spots have equal design scores results in the emergent hot spot overtaking the old one. In particular, Fig. 19 shows A = 1 and B = 2. Conceptually, increasing B beyond this value results in smaller and smaller links being favored. This makes sense both mathematically and physically. Mathematically, the denominator of the modified design score equation becomes unboundedly large as B increases, and the denominator of the design score equation increases accordingly much faster than the isotropy terms in the numerator could limit it. This continues until an inevitable singularity occurs due to these now smaller links trying and being unable to reach the boundaries of the workspace of the finger. This sets the design score to zero where the emergent hot spot is when B = 3.75, ultimately creating a cold spot beyond which any link length combinations are unacceptable. In terms of the physical prototype and finding this result intuitively and through experiment, this corresponds to the situation in which smaller link lengths were favored much more compared to the tactile smoothness of the mechanism's movement. The links would become smaller and smaller until they could no longer physically reach where the hand of a user attempts to move. This, along with any smaller link combinations, limits the motion of the hand of the user, and the set of link lengths would be deemed unacceptable. Thus, the shortest acceptable set of link lengths would be reached both mathematically through the algorithm and physically through experimentation.

[00107] In addition to causing a new hot spot to emerge, the addition of weight factors to the design score can also cause the creation of a sizeable cold spot within the plane. This rectangular cold spot is in the region where L ₁ is small and L3 is large as is seen in the subplots with B > 2 shown in Fig. 19. When A = 1 and B = 1 , the plane of design score values can be seen as relatively flat with no deep wells. The most interesting feature is located at the area surrounding the hot spot with large L ₁ and small L3. In this lower-right region of the plot, the design score changes quickly over small changes in link length. However, when link length is weighted more heavily, the topology of the map changes. In addition to there being a greater number of contour lines in the area nearby the hot spot, the aforementioned cold spot is created that shows there is a steep drop in design score. Physically, one interpretation of this cold spot is that any increase in the mechanism isotropy terms caused by decreasing L ₁ and increasing L3 (which is one direction of increasing design score until the cold spot is suddenly reached) is quickly negated and overwhelmed by the benefit to link length reduction gained from decreasing L ₁ . This is due to the fact that the first link is the longest link in the mechanism. Thus, the first link influences the denominator of the design score equation much more than the other two links.

[00108] Another striking graphical effect worth interpreting is the rearrangement of the contour lines with increasing weight factor B. For B = 1 , the contour lines indicate that design score increases with increasing L ₁ and decreasing L3. This reflects how physically, a large L ₁ results in smooth mechanism movement, increasing the isotropy terms in the design score equation. The same pattern holds for the case with B = 2. However, as the new hot spot emerges with B > 3.75, the contour lines rearrange to indicate that the design score increases as both L ₁ and L3 decrease. This new direction of increase is visually almost perpendicular with the one from the cases where link length was not prioritized as heavily. At this point, the weight factors A and B are adjusted such that the plot suggests that the way to achieve the optimal design is to decrease both link lengths as much as possible without the mechanism reaching a singularity or failing to cover the entire finger workspace. This is in direct agreement with the main design objective used for an experimental determination of the optimal link lengths. Therefore, the direction of increasing design score as shown by the contour lines is an important feature in the plots that can be used to judge whether or not the design score equation in the algorithm is in agreement with the strategy driving the experimental optimization.

[00109] Although the above discussion presents a specific case in which a 3R planar mechanism is the mechanism of interest, the methodology behind the optimization could be applied to any other type of mechanism in which an end effector controls a body part that interfaces with mechanism at a fixed point, provided that the kinematics and inverse kinematics are solvable. The weight factors used in an optimization in accordance with some embodiments of the invention make the design score equation flexible to fit the priorities of the designers.

[00110] Improvements that can be made to the optimization method in accordance with various embodiments of the invention include, but are not limited to, optimizing the mechanism towards specific regions within the desired workspace. For instance, statistics may show that a hand spends most of the time during activities of daily life in certain clenched positions. These positions can be mapped to a region within the overall workspace of the fingers, and the design score equation and weight factors can be modified to prioritize kinematic performance in these particular areas of the workspace in accordance with a number of embodiments. The same principle can be applied to optimize performance in workspaces areas where stroke patients with spasticity have particular difficulty moving in accordance with a few particular embodiments. [00111] Although the above discusses optimization processes using a design score in accordance with a few embodiments of the invention. Some other embodiments of the invention can use optimization processes that add, remove, combine, and/or modify some of the discussed steps and equations.

Control System

[00112] As a result of the third joint being passive and the nature of how the finger interfaces with the third link in a hand exoskeleton manipulator in accordance with some embodiments of the invention, the force exerted by the user is generally in line with the third link. The bending beam load sensor link in Link 2 of each of the finger mechanisms can sense the component of the end effector force that is perpendicular to the sensor. Using the angle information from the encoders, the measured force can be used to determine the force at the end effector with respect to the ground frame of the device. The joint angles along with the end effector force of a 3R planar mechanism in accordance with an embodiment of the invention are shown in Fig. 20. The planes of the linkages as well as an end effector force vector measured by the sensor on link 2 are shown in Fig. 21 . The end effector force vector along with the force component measured by the sensor on Link 2 are shown in Fig. 22.

[00113] In accordance with some of these embodiments, the Fs read by the sensor can be expressed in reference frame 3 as:

Therefore, the force at the end effector of the finger manipulator as expressed in frame 3 is:

[00114] For the thumb 3R planar mechanism, which has the sensor on the first link, the force at the end effector as expressed in frame

[00115] Based on the joint angles in Fig. 20, the rotation matrix relating reference frame 3 and the ground frame of a hand exoskeleton in accordance with this embodiment of the invention is:

Where C123 = cos(0i + Θ2 + Θ3), and S123 =( Θ1 + Θ2 + Θ3). In general, the end effector force in the hand exoskeleton ground frame is then:

This is the force that is taken to be the input to an admittance control algorithm in accordance with some embodiments of the invention.

[00116] As discussed above, Joints 1 and 2 are actuated while Joint 3 is passive. Therefore, a control algorithm in accordance with many of these embodiments controls a 3R planar mechanism and a 2R planar mechanism corresponding to the first and second links. As such, the position of the end effector is controlled to be within a circle (or, due to the links' physical constraints in accordance with numerous embodiments, a semicircle) with radius equal to the length of the third link L3. Although the angle of the third link is not controllable, control systems in accordance with some embodiments measure the angle and use the angle to calculate the position of the end effector. This is useful for determining the end position of the fingertip during rehabilitation exercises. The forward kinematics and Jacobian of the 2R planar mechanism are given by:

To find the joint torques of a 3R planar mechanism, the force at the end effector can be multiplied by the transpose of the Jacobian matrix in accordance with many embodiments as shown in this equation:

[00117] The admittance control relation in accordance with some embodiments uses these torques to calculate the desired joint angles for the next time step. The relation used is:

Where k _pa , kia, and kda are the admittance control gains for the proportional, integral, and derivative terms, respectively. The desired joint angles are then used as the set points in a PID control scheme given by the equation in accordance with a number of embodiments:

where vim is the voltage output sent to the motor for joint I; e _i is the error for joint I; and kpi, kn, and kdi are the PID gains for the ith joint. Note that in the above equation, the usual derivative term, is replaced with the gain kdi multiplied by the negative derivative

of the voltage. This is done to prevent derivative kickback.

[00118] A block diagram of components of a hand exoskeleton manipulator and a control system in accordance with some embodiments of the invention is shown in Fig. 23. As can be seen in Fig. 23, the user of the exoskeleton first exerts a force on the end effector. The force sensors, the second link for the finger mechanisms and the first link for the thumb mechanism, read a component of that force in accordance with the angles of the links. The force sensors output an analog voltage to a control system device. The signal goes through a low pass filter and is amplified and translated into a force. Using the previous equations, the force is transformed to the ground frame and the joint torques are calculated. These torques are used as the input to an admittance control block, and the new set point angles are calculated. A PID controller can then output voltages to the motors to bring the manipulator to this desired position.

[00119] An initial determination of the gains for the admittance relation and PID controller may be performed in the following manner in accordance with some embodiments of the invention. First, the gains for the PID controller are set. This is done in order to create a desired response of the motors, end effector, and/or joints to changing angle set points. The qualitative objective is to create a response that results in movement that is subjectively comfortable to the human hand. Because steady state error is less of an issue in a system in accordance with these embodiments, a PD controller can be used, and the integral gains for both joints were set to zero.

[00120] During testing, the proportional gain may be increased until the system response begins to oscillate. Following this, the derivative gains are increased from zero until the system settles within a desired time. It was observed during tuning trials that for a given set of gains, the system response can vary in response to steps of various sizes, indicating nonlinearity in the system. Therefore, the PID gains can be tuned using set point changes of approximately 10 degrees, which are still higher than those that the admittance control algorithm is expected to attempt.

[00121] The admittance relations were the next focus. These parameters can influence the trajectory of the end effector that the control system plans at each time step in response to the force exerted at the end effector by the user of the hand exoskeleton. For initial testing of the control algorithm, only a proportional gain of non-zero is used to experiment with the effects of the integral and derivative terms on trajectory. This made the change in the motor's angles directly proportional to the force at the end effector, giving control of a straightforward stiffness in the manipulator. The values for both the admittance relation and PID controller gains used for initial testing of a hand exoskeleton manipulator in accordance with an embodiment of the invention are shown in the following table:

[00122] The above is a description of a control system for a hand exoskeleton manipulator in accordance with some embodiments of the invention. However, other control systems may modify, add, remove, and/or combine steps and/or components in accordance with various other embodiments of the invention.

[00123] Various embodiments of the mechanical design and base control algorithm of a dexterous hand exoskeleton for use in stroke patient rehabilitation are described above. The exoskeleton can control the fingers and thumb with three 3R planar mechanisms. In accordance with many embodiments, the base positions of these mechanisms can be adjusted and the end effector interfaces with the user are designed to be switchable. Thus, the exoskeleton in accordance with some embodiments of the invention has the ability to manipulate different combinations of fingers and aid the user with a number of different types of grasps during rehabilitation exercises. A brute force kinematic optimization method may be used as a tool to determine the manipulator link lengths that allow the device to effectively accommodate a wide range of hand sizes. The three 3R planar mechanisms are actuated by a remote motor pack through a system of capstans and Bowden cables. Each manipulator operates with an admittance control scheme and assists the user based on the input force to the end effector. [00124] Some further embodiments may fine-tune the gains for the admittance controller based on the overall control strategy. One factor to consider is how much assistance the user requires from the exoskeleton during certain motions or grasps. This opens up the possibility of a heuristic control system to learn which gain settings are appropriate and/or adjust these gains in real time in accordance with a number of embodiments.

[00125] Some further embodiments may include an improved optimization algorithm. Adjustments can be made to the algorithm based on new data that might become available in the future. For instance, if the relevant data becomes available, the algorithm, instead of optimizing manipulator kinematics for the entire desired workspace, could prioritize performance in sub-areas of the workspace that fingers are found to spend the most time in during activities of daily living, or those regions that stroke patients with spasticity have the most difficulty moving in. An optimization algorithm accounting for these data would aid in the mechanical design of any future iterations of the manipulators. A number of these further embodiments could also experiment with sensors of different types and smaller sizes. This could allow the force sensing to be done more proximally to the finger interface.

[00126] Sensors from cell phone technology may also be incorporated into the main body of the exoskeleton hand in accordance with many embodiments to sense its absolute orientation relative to other components of an entire upper-limb exoskeleton.

[00127] Although the present invention has been described in certain specific aspects, many additional modifications and variations would be apparent to those skilled in the art. It is therefore to be understood that the present invention may be practiced otherwise than specifically described, including various changes in the implementation, without departing from the scope and spirit of the present invention. Thus, embodiments of the present invention should be considered in all respects as illustrative and not restrictive.