

Design and Testing of a Novel, High-Performance

Two DoF Prosthetic Wrist
Revanth Damerla, Kevin Rice, Daniel Rubio-Ejchel, Maurice Miro, Enrico Braucher, Juliet Foote, Issam Bourai,
Aaryan Singhal, Kang Yang, Hongju Guang, Vasil Iakimovitch, Evelyn Sorgenfrei, Shorya Awtar

Abstract— Extrinsically powered prosthetic wrists have the Unfortunately, rejection rates for extrinsically powered
potential to offer significant improvements to the functionality and prosthetic hands without a prosthetic wrist have remained
dexterity of a prosthetic hand. They can also reduce a user’s consistently high (>20%) over the past 20 years [18]–[22].
overreliance on their intact limb and help prevent injury from
overuse of upper limb (both intact and residual) and trunk joints. Furthermore, studies have documented that unilateral upper
Despite these potential advantages, there are very few prosthetic limb myoelectric prosthesis users (i.e. having one intact upper
wrist options that are commercially available and these devices are limb and using a prosthetic hand without a prosthetic wrist on
not commonly used by prosthetic hand users due to several factors the amputated limb) are over reliant on their intact upper limb
including inadequate performance specifications. In this paper, we [23], [24]. One study has shown that 75-94% of upper limb use
first seek to establish the target specifications for a prosthetic wrist by a unilateral amputee was with their intact limb [24]. An
suitable for both median men and women. We then complete a
comprehensive review of the state-of-the-art in extrinsically additional study found that the median use of the intact limb for
powered prosthetic wrists in the research, commercial, and patent individuals with unilateral upper limb absence was 79% while
literature. This review determines that no existing prosthetic wrist the median use of the dominant hand in individuals with two
meets the target specifications due to the presence of actuators and intact upper limbs was 52% [23]. A large proportion of
transmissions that do not offer sufficient torque density, power amputees (>50%) have also reported significant pain and
density, and specific power. In order to address this challenge and injuries to both their intact and residual limbs and other parts of
produce a prosthesis that achieves target specifications, we next
review the performance of existing actuators and transmissions the body including the trunk, neck, and back [25]–[28]. The
and determine that Brushless DC motors with planetary gearboxes pain and injuries can come from several sources including
and ball screws offer the best potential to achieve the target phantom limb pain or overuse of joints due to overreliance of
specifications. We then present the design of a novel two Degree of the intact limb or compensatory motions of various bodily joints
Freedom parallel kinematic prosthetic wrist that incorporates this while completing tasks; injuries can include arthritis, joint
actuator-transmission combination. This first iteration of the degradation, muscle injuries, and tendonitis [25]–[28]. These
proposed prosthetic wrist meets the target torque, speed, and
weight but does not meet the target dimensions or range of motion high rejection rates, the overreliance on the intact limb, and
yet. We propose design improvements in subsequent iterations prevalence of pain and injuries in amputees are significant
that could lead to a prosthetic wrist that meets all the target because they demonstrate that user needs are not adequately
specifications of torque, speed, weight, and volume. met with current prosthetic hands only (i.e. without a wrist)
despite numerous recent technological advancements.
Index Terms— Actuators, biomechatronics, mechanical In addition, when both non-users (i.e. people who have
transmissions, physiology, prosthetics
rejected extrinsically powered prosthetic hands) and users with
transradial (at the forearm) amputations were asked how
prosthetic hands could be improved, several answers were
I. INTRODUCTION AND MOTIVATION
consistently given over the past 20 years including reduced

E XTRINSICALLY powered upper limb prostheses,

including prosthetic hands [1]–[8], wrists [9], elbows, and
shoulders [10]–[13] have the potential to help the 41,000
weight [19]–[22], [29]–[31], increased comfort of the interface
with the user (e.g. socket or harness) [19]–[22], [29], [32],
improved controllability and better ease of use [21], [22], [29],
amputees missing more than their fingers in the United States [30], [32]–[34], increased durability [20]–[22], [29], [30], [32],
[14] and many more worldwide [15]–[17] overcome the and better functionality to enable the user to complete desired
challenges associated with upper limb loss. These challenges ADLs [19]–[22], [29], [34]. Better functionality is related to
include efficiently completing activities of daily living (ADLs) several specific improvements including increased dexterity
and maintaining a sufficiently high degree of independence. For (both of the hand and wrist) [22], [29], increased hand strength
those with limb loss at the wrist (i.e. wrist disarticulation) or [32], and increased hand speed [31]. Dexterity requires both the
more proximally, the prosthesis must ideally comprise an end hand and wrist to complete small, precise actions (colloquially
effector that approximates the functionality of the lost hand as called “fine-motor skills”) and accurately achieve the desired
well as the wrist to allow the user to complete ADLs. hand posture or grasp (i.e. have the necessary independently

R
evanth Damerla was supported by a National Science Foundation Graduate contributed equally to this work.) (Corresponding author: Revanth Damerla.)
Research Fellowship during this research. (Kevin Rice, Daniel Rubio-Ejchel, All authors are with the Precision Systems Design Laboratory, Department
Maurice Miro, Enrico Braucher, Juliet Foote, and Issam Bourai contributed of Mechanical Engineering, University of Michigan, Ann Arbor, MI 48109
equally to this work.) (Hongju Guang, Vasil Iakimovitch, and Evelyn Sorgenfrei USA (e-mail: [email protected]).

© 2022 IEEE
IEEE Transactions on Medical Robotics and Bionics 2

controlled Degrees of Freedom (DoFs)).

The prosthetic wrist is a crucial component to addressing the
user need of better functionality, particularly increased
dexterity, in a prosthetic hand. Just as the human wrist and
forearm provide critical articulation that adjusts the orientation
of the hand and help it achieve a desired hand posture, the
prosthetic wrist can provide the same for a prosthetic hand.
Thus, the absence of a prosthetic wrist for prosthetic hand users,
even a passively flexible or manually adjustable wrist, can have
significant repercussions. Several studies have demonstrated
that completing ADLs without a wrist requires significant
compensatory motions of both the shoulder and trunk (e.g.
leaning to the side) both for prosthetic hands without a
prosthetic wrist [35]–[38] and for intact upper limbs with a
fused (via a brace) wrist [38]–[40]. As discussed above, Fig. 1 DoFs of the Human Wrist and Forearm
compensatory motions can lead to overuse of joints over time
that can lead to injuries and pain. prosthetic hand despite the above-described potential benefits.
Several studies have also investigated whether a passively In this paper, we seek to first understand why commercially
flexible [41]–[43] or manually adjustable [43], [44] prosthetic available prosthetic wrists have inadequate specifications and
wrist can offer a performance improvement versus no wrist to a secondly to begin to address this challenge with a novel two
myoelectric prosthetic hand user. These studies have not DoF prosthetic wrist capable of F/E and Radial/Ulnar Deviation
demonstrated improved performance across all tasks. This is (R/U). While the ultimate objective is to provide a prosthesis
not surprising given that these prosthetic wrists are not with all three DoFs (which is necessary to provide sufficient
extrinsically powered and therefore cannot be actively (e.g. dexterity), the presented prosthetic wrist is an important
myoelectrically) controlled. However, these studies and one intermediate step toward achieving this objective. We begin by
survey have consistently found that prosthesis users prefer the identifying the target specifications for a prosthetic wrist that
functionality and added dexterity of these prosthetic wrists over can be used by a median man or woman, which are based on the
no wrist [41]–[45]; one study also found that users would have performance of the human wrist (Section II). We then review
further preferred an extrinsically powered prosthetic wrist [41]. the specifications of existing prosthetic wrists and show that
One additional study has investigated the importance of a none of them meet the target specifications (Section III). It is
prosthetic wrist by constraining the number of DoFs of a human shown that the current gap is mainly due to limitations of
hand and wrist in subjects with intact limbs [39]. This study existing actuators and transmissions, which cannot provide
demonstrated that a two DoF wrist capable of Pronation/ adequate torque density, power density, and specific power to
Supination (P/S) and Flexion/Extension (F/E) (see Fig. 1) with achieve the target specifications (Sections III and IV). This
a 1 DoF gripper performed approximately as well as an intact motivates the design of a new prosthetic wrist that overcomes
23 DoF human hand with a 1 DoF P/S wrist in completing the the limitations of existing prosthetic wrists through the
Southampton Hand Assessment Protocol (SHAP) tasks [39]. selection of actuators, transmissions, and a mechanism that can
These findings demonstrate the importance of an achieve the target prosthetic wrist specifications. Next, we
extrinsically powered prosthetic wrist, which can provide the review the capabilities of existing actuators and transmissions
significantly better dexterity users desire in their prosthetic to identify that Brushless DC Motors with planetary gearboxes
hand, reducing reliance on their intact limb and compensatory and ball screws offer the best potential to achieve the target
motions. This can improve their ability to perform ADLs and prosthetic wrist specifications (Section IV). In Section V, we
also lower the risk of injuries from overuse of joints. However, present the design of the novel two DoF parallel kinematic
most commercially available prosthetic hands are not sold with prosthetic wrist. The presented two DoF prosthetic wrist
an accompanying prosthetic wrist [3], [8]. There are a few incorporates the above actuator-transmission combination and
separately sold options for prosthetic wrists [10]–[12], [46]– can achieve the target prosthetic wrist performance and weight
[51] that typically offer only one (either P/S or F/E) DoF. As but is slightly larger than the target dimensions and does not
discussed above, a prosthetic wrist with more than one DoF meet the target range of motion. We also suggest design
provides significantly improved performance over just one DoF improvements for how a future iteration of this prosthesis can
and a prosthesis with all three wrist DoFs is necessary to achieve all target specifications. Finally, Section VI presents
provide sufficient dexterity (see Section IIC). Furthermore, the evaluation of a fabricated prototype of the prosthetic wrist
existing prosthetic wrists are relatively heavy and do not offer to report how well it achieves the desired target specifications.
torque and speed outputs that are comparable with the human
wrist and forearm (i.e. inadequate specifications – see Section II. HUMAN WRIST BACKGROUND
III). Finally, these options also entail additional cost and A. Architecture of the Human Wrist and Forearm
complexity. Given this, most prosthetic hand users decide to not The human wrist consists of eight small carpal bones that
include an extrinsically powered prosthetic wrist with their
IEEE Transactions on Medical Robotics and Bionics 3

connect the distal ends of the two bones of the forearm – the TABLE I
HAND AND FOREARM WEIGHT AND VOLUME FOR MEN AND WOMEN
radius and ulna – to the proximal ends of the metacarpals in the
hand [52]. Together with the radius, ulna, and metacarpals, the Dimensions Men Women
carpal bones provide two DoFs through a complex of joints: 1. % Body Weight 0.63 [76]–[79] 0.53 [77], [79]

Hand
Flexion and Extension (F/E) and 2. Radial and Ulnar Deviation Mass of Hand for Median US
540 380
Adult by Weight (g)
(R/U) (Fig. 1). The radius and ulna also provide a third DoF –
Volume (cm3) 363 [76], [78]
Pronation and Supination (P/S) – by crossing over one another
% Body Weight 1.66 [76]–[79] 1.48 [77], [79]
along the length of the forearm [53].

Forearm
Mass of Forearm for Median
Each of these DoFs is actuated by muscles and tendons US Adult by Weight (g)
1420 1060
housed in the forearm. F/E and R/U are primarily provided by Volume (cm3) 909 [76], [78]
five muscles that originate near the elbow and insert near the
proximal ends of the metacarpals [54]. These muscles are TABLE II
located in the superficial layers of the posterior and anterior MEAN HAND, WRIST, AND FOREARM DIMENSIONS FOR MEN AND WOMEN
compartments of the forearm (i.e. near the surface of the top and Dimensions Men Women
underside of the forearm). The location of these muscles and Wrist Thickness (mm) [57], [58] 43.0 37.0
tendons produces the highest possible moment arms about the Wrist Width (mm) [55]–[59] 63.1 56.1
wrist and therefore enables higher joint torques. Actuation for Wrist Circumference (mm) [57]–[61] 172.4 149.5
P/S on the other hand is primarily provided by three muscles Flexed Forearm Circumference (mm) [59], [61] 307.0 259.1
Radial-Stylion Length (mm) [59], [61] 268.6 242.5
that originate on the ulna and elbow and insert onto the radius,
allowing the two forearm bones to rotate relative to one another TABLE III
[54]. Two additional muscles assist in actuation of this DoF and MEAN WRIST AND FOREARM RANGE OF MOTION
help stabilize the forearm [54]. Degree of Freedom Maximum RoM (°) Functional RoM (°)
B. Human Wrist and Forearm Weight and Dimensions Pronation 83 [62]–[66] 61 [68]–[71]
The mean weight and volume of the hand and forearm for Supination 100 [62]–[66] 75 [68]–[71]
Flexion 76 [62]–[64], [66], [67] 54 [71]–[74]
men and women are shown in Table I and are adapted from [8].
Extension 73 [62]–[64], [66], [67] 48 [72]–[75]
Given that a transradial (at the forearm) amputee will have some Radial Deviation 25 [62]–[64], [66], [67] 22 [73]–[75]
of their forearm intact, less than half of the weight of the Ulnar Deviation 45 [62]–[64], [66], [67] 38 [71]–[74]
forearm is a logical target for the weight of the prosthesis.
Dimensions of the wrist and forearm are listed in Table II, generally not required (Functional RoM in Table III) [68]–[75].
which were obtained from several studies [55]–[61] that Since these studies used different ADLs in determining
measured these dimensions for U.S. adults with average ages Functional RoM, certain studies found significantly lower
under 30 or people serving in the U.S Armed Forces. As noted values in certain DoFs. These outlier values were removed in
previously [8], this may lead to values that reflect people who order to ensure a representative maximum Functional RoM is
are younger and more muscular than the typical prosthesis user. listed. Table III illustrates that Functional RoM is substantially
In studies that listed the median height and weight of study lower than Maximum RoM in both F/E and P/S. The Functional
participants [57]–[60], the median weight of participants in the RoM also demonstrates the utility of each DoF in completing
studies was in some cases over 10kg less than the median U.S. ADLs. A prosthesis with only one or two DoFs will therefore
adult while the median heights of participants were within 4 cm likely not provide adequate dexterity. R/U in particular is
of median U.S. adults. This difference in height is significantly frequently left out of prostheses; while the human wrist has less
smaller than the difference in weight. Thus, length dimensions RoM in this DoF, the Functional RoM demonstrates that
in Table II are more likely to be representative of the general articulation in R/U is still valuable.
U.S. population than the width, thickness, and circumference The maximum joint torque in each DoF is reported for both
dimensions. However, the wrist is relatively bony and generally men and women in Table IV. Studies that measure these values
lacks significant muscle mass or fatty tissue, meaning these typically measured maximum joint torque in several postures
dimensions may still be representative of the general U.S. (e.g. wrist fully pronated during a P/S torque measurement) or
population despite the difference in weight with participants of at several joint speeds (e.g. P/S torque measurement at both 0
the studies. They can therefore still help serve as target °/s and 30 °/s in P/S). Therefore, two separate maximum values
dimensions for a prosthesis. are listed for both men and women. True Maximum refers to
C. Human Wrist Performance the maximum value measured in each study across all postures
or joint speeds while Mean Maximum uses the mean across all
Wrist Range of Motion (RoM) in its 3 DoFs is a critical
postures or speeds. True Maximum is therefore always higher
component of the dexterity it provides and is shown in Table
than Mean Maximum. It is likely acceptable to use the Mean
III. While the human wrist can reach certain extreme values in
Maximum value in Table IV as a target for a prosthesis as the
each DoF (denoted as Maximum RoM in Table III) [62]–[67],
True Maximum reflects a single orientation or speed that is
several studies have measured wrist RoM during completion of
unlikely to consistently be used in practice.
ADLs and have demonstrated that the full RoM of the wrist is
IEEE Transactions on Medical Robotics and Bionics 4

TABLE IV
MAXIMUM WRIST JOINT TORQUES FOR MEN AND WOMEN
Degree of Men Women
Freedom True Maximum Mean Maximum True Maximum Mean Maximum
Pronation 10.3 [58], [84]–[86] 9.0 [58], [84]–[86] 5.1 [58], [85], [86] 4.5 [58], [85], [86]
Supination 10.8 [58], [84]–[86] 9.5 [58], [84]–[86] 5.3 [58], [85], [86] 4.6 [58], [85], [86]
Flexion 14.6 [57], [58], [87], [88] 12.7 [57], [58], [87], [88] 9.7 [58], [88] 8.8 [58], [88]
Extension 9.3 [57], [58], [87], [88] 7.9 [57], [58], [87], [88] 6.8 [58], [88] 5.8 [58], [88]
Radial 14.3 [58] 13.0 [58] 8.8 [58] 8.2 [58]
Deviation
Ulnar 13.5 [58] 12.4 [58] 8.8 [58] 8.0 [58]
Deviation

The mean maximum wrist joint speed in each DoF is reported TABLE V
MEAN MAXIMUM WRIST JOINT SPEEDS
in Table V. Unfortunately, relatively few studies have measured
Mean During
these values. However, three different values are reported for Degree of Max During
Peak (rad/s) Practical
each DoF when available. Peak refers to the maximum Freedom Tennis (rad/s)
Tasks (rad/s)
instantaneous joint speed that can be measured. Two studies Pronation 38 [62] 19 [81]
have provided measurements for these values [62], [80]. Supination 33 [62] 5 [81]
However, the peak value is substantially higher than what is Flexion 27 [62], [80] 17 [81] 1.7 [82], [83]
likely to be used in practice. One study measured the maximum Extension 26 [62], [80] 8 [81]
joint speeds in tennis players, which requires relatively extreme Radial Deviation 10 [62], [80] 3 [81] 1.7 [82]
Ulnar Deviation 11 [62], [80] 15 [81]
maximum joint speeds in all three DoFs [81]. Finally, two
studies have measured the joint speeds during practical tasks TABLE VI
[82], [83], with one measuring mean joint speeds across many TARGET PROSTHETIC WRIST SPECIFICATIONS
professions [83]. These values provide a reasonable, but lower Weight or
Value
bound on the speeds necessary to effectively perform ADLs. Dimension
Two studies have recommended that the maximum joint speed DoFs F/E and R/U
for Flexion/Extension of the fingers of a prosthetic hand be 3.0- Weight (g) 260-370
3.5 rad/s [1], [2], which can also serve as an upper limit for Width (mm) 55-60
maximum joint speeds in a prosthetic wrist. Beyond these Thickness (mm) 35-40
Length (mm) 70-100
speeds, controllability can become an issue for prosthetic hand
users [3]. Performance F/E R/U
RoM (°) 55/55 25/45
D. Target Prosthesis Specifications
Joint Torque (Nm) 8-12 8-12
Table VI describes the target specifications for the prosthetic Joint Speed (rad/s) 2-3.5 2-3.5
wrist design presented below, which are informed by the above
values for the human wrist and forearm and the motivation to Maximum joint torque for both men and women. This joint
produce a prosthesis that provides sufficient performance to torque should be adequate for all ADLs except those requiring
address the user needs for both median men and women. The more extreme strength such as carrying very heavy objects. A
prosthesis described in this paper will only provide F/E and R/U joint speed target of 2-3.5 rad/s was set to exceed mean speeds
and a future iteration will include P/S. The target weight for the during ADLs but not exceed what is likely to be controllable for
prosthesis was set to 260-370 g, 25-35% of the median female the user. As with the human wrist, the target torque and speed
wrist to ensure the inclusion of P/S would not lead to an performance are not required simultaneously. Instead, high
infeasible weight for the intended users. A target width and torque would be required at low speed (i.e. near stall) and high
thickness of 55-60 mm and 35-40 mm respectively were speed would be required while holding relatively small weights.
initially chosen. Although these values can be larger than for a
median woman, it enables the largest possible space for III. REVIEW OF EXISTING PROSTHETIC WRISTS
actuators, transmissions, and other components and is unlikely
In order to develop a prosthesis that meets the above target
to prevent a user from completing most ADLs. For example, the
specifications, it is crucial to understand the specifications of
median woman’s palm thickness at the thenar pad is much
existing prostheses and determine whether any have achieved
thicker – 51.7 mm [55]. Finally, a target length of 70-100 mm
the target specifications. We have therefore conducted a
was chosen because it is about 30-40% of a median woman’s
comprehensive review of the research, patent, and commercial
radial-stylion length. This length enables future inclusion of P/S
literature. This review builds upon the findings of a previous
while still being suitable for transradial amputees.
paper that has reviewed extrinsically powered prosthetic wrists
RoM in both DoFs was informed by the Functional RoM,
[9] but differs significantly by specifically focusing on each of
where 25/45 for R/U denotes a 25° RoM in Radial Deviation
mechanisms, actuators, and transmissions. These components
and a 45° RoM in Ulnar Deviation, leading to a total of a 70°
of an extrinsically powered prosthesis account for the majority
RoM in R/U. A joint torque target of 8-12 Nm was set in both
of the weight and size and provide its performance.
DoFs to provide a sufficiently large percentage of the Mean
IEEE Transactions on Medical Robotics and Bionics 5

A. Methods
For this review, a similar process to that used in [8] was used.
Extrinsically powered prosthetic wrists described in English
and developed between 2000-2021, a period that adequately
covers most modern innovations in the field, were identified
using several search engines and the following search terms:
(“prosthetic hand”, “prosthetic gripper”, “upper limb
prosthesis”, “robotic hand”, “prosthetic wrist”) by themselves
and in conjunction with the terms (“powered”, “extrinsically
powered”, “active”); the references cited by each source were
also reviewed to ensure this review examined as many
prostheses as possible. The measured specifications of each
prosthesis were recorded. In cases where certain values could Fig. 2 Examples of Prosthetic Wrists: a. Motion Control Standard Electric Wrist
not be found, the corresponding authors, companies, or creators Rotator [46], b. Modular Prosthetic Limb v3 [13], c. Unnamed [98], d.
were contacted to try to obtain the missing values. If the design Unnamed [99]
was never built or if no specifications could be found, the
prosthesis was removed from the review as it could not be transhumeral or more proximal amputations). While this weight
adequately compared. Prostheses were also removed from the distribution can be altered for a prosthesis with 3 DoFs, 50-75g
review if newer versions existed, leaving a total of 21 for each actuator and transmission is a reasonable target for the
prostheses compiled here. 2 DoF prosthesis capable of F/E and R/U described in this work.
The specifications of each prosthesis are compiled in Table This weight also leaves a reasonable weight for the other
VII. In several cases, only theoretical values were reported for mechanical and structural components of the prosthesis.
a prosthesis. These values are not reported in the table because A final consideration is the type of mechanism used in the
they would imply an efficiency of 100%. Given the high prosthesis – serial kinematic, parallel kinematic, or a
reduction ratios often employed in these prostheses, this combination of the two. In the human wrist, F/E and R/U are
assumption is not practical. Losses due to friction, viscous provided by a parallel kinematic mechanism (PKM) and are
damping, inertia, and other factors can substantially impact connected in serial with P/S. This takes advantage of several
performance. For example, the theoretical maximum torque aspects of both types of mechanisms. Serial kinematic
output for the prosthesis reported in [89] is 0.584 Nm. However, mechanisms (SKMs) can be simpler, more modular, and have
the measured maximum torque is 0.0596 Nm, meaning an greater ranges of motion than the parallel kinematic options
efficiency of approximately 10%. [92]–[94]. However, if both F/E and R/U are provided through
An interesting consideration is the total weight of muscle a serial kinematic approach, the axes of rotation for these DoFs
used to actuate the 3 DoFs of the wrist. These values were may not intersect as they do in the human wrist [13]. Currently,
calculated from the measured volume of each muscle no study has been conducted to demonstrate whether non-
responsible for actuating each DoF from [90]. Unfortunately, intersecting F/E and R/U axes are unintuitive for prosthesis
this study, which measured muscle volumes for both living men users or lead to any loss in performance. PKMs on the other
and women, did not list mean muscle volume for men and hand can have ground-mounted actuators, can be more compact
women separately but provided a single average. However, a and lightweight, and potentially have higher speeds, torques,
different study could not be found providing these volumes in and stiffnesses [92]–[94]. In a prosthesis, the benefits between
living men and women (and not cadavers) separately. The these two types of mechanisms must be compared while also
volumes were converted to weight assuming a muscle density accounting for the chosen actuator and transmission
of 1.037 g/cm3 [91]. Since F/E and R/U are actuated by the same combination. Given the strict spatial constraints of a prosthesis,
group of muscles, the weight for their actuation was considered specific actuator and transmission combinations may only be
together. The muscle weight for P/S includes the brachioradialis feasible with either a SKM or PKM.
muscle, which also aids in flexion of the elbow. The resultant B. Prosthesis Review
weights can serve as a helpful benchmark for the maximum Eight of the reviewed prostheses provide a single DoF,
target weight of the actuators and transmissions used in a including four of the five commercially available prostheses.
prosthetic wrist; the standard deviation can then be used to These prostheses were intended for either P/S [46]–[49], [89],
adjust targets for prostheses meant specifically for either men [95] (Fig. 2a) or F/E [50], [51], [96], [97]. Of these eight, at
or women. Ideally however, these components should weigh least four (but likely more) of the eight prostheses use a Brushed
even less than human muscle to ensure the prosthesis will not or Brushless DC (BDC or BLDC) Motor connected to a
be too heavy for the user. For a prosthesis meant for both men gearbox, whose output directly drives a wrist joint [50], [51],
and women with transradial amputations, the actuators and [89], [96], [97]. At least one additional prosthesis uses an
transmissions for F/E and R/U each should weigh 50-75g while ultrasonic motor instead [95]. A common theme among these
a P/S module could weigh 100-150g (assuming elbow flexion prostheses is that they often provide at least a functional range
can be provided by the residual limb or by actuators and of motion, with several P/S prostheses providing a full 360°+
transmissions housed in the upper arm for users with
IEEE Transactions on Medical Robotics and Bionics 6

TABLE VII
PROSTHESIS SPECIFICATIONS
Width ×
Serial or Length Weight Max Joint Max Joint Actuator
Prosthesis Name Year #/Type of DoFs RoM (°) Thickness Actuator–Transmission Pathway (Weight (g))
Parallel (mm) (g) Torque (Nm) Speed (rad/s) Type
(mm)
9.0-10.8 – P/S,
Median Male Wrist 63.1 × 43.0 280-570* 7.9-14.6 – F/E,
83/100 – P/S, 38/33 – P/S,
3 (F/E, R/U, P/S) Serial and 12.4-14.3 – R/U Human P/S – Muscle (151±75)-Tendon
76/73 – F/E, 27/26 – F/E,
Parallel 4.5-5.3 – P/S, Muscle F/E+R/U – Muscle(153±69)-Tendon
25/45 – R/U 10/11 R/U
Median Female Wrist 56.1 × 37.0 210-420* 5.8-9.7 – F/E,
8.0-8.8 – R/U
55/55 – F/E, 55-60 × 2-3.5 – F/E,
Target Prosthetic Wrist 2 (F/E, R/U) 70-100 260-370 8-12 – F/E, R/U
25/45 R/U 35-40 R/U
F/E, R/U – Maxon ECX Speed 13L HP (33.7+) – Maxon
48/33 – F/E, 8.2 – F/E, 4.2 – F/E,
This Design 2021 2 (F/E, R/U) Parallel 66 × 52 131.4 320# BLDC GPX 13 Speed 25:1 (26+) – 1:1 Spur Gears (8) - Ball
33/25 – R/U 8.4 – R/U R/U
Screw + Nut (10) – S-S Link (11)
Motion Control Standard
Electric Wrist Rotator 1 (P/S) N/A 47 × 47 70 143 1.7 2.9
(comm.) [46]
Motion Control Powered
46.74 × BLDC – 2-Stage Friction Planetary Drive – 32:3 Evoloid
Flexion Wrist (comm.) [50], 1 (F/E) 86/67 N/A 66 258.55 2.3 3.1 BLDC
46.74 Gear Drive
[51]
Ottobock Electric Wrist
1 (P/S) 360+ N/A 96 1.78
Rotator (comm.) [47]
Ӧssur i-Limb Wrist (comm.)
1 – (P/S) N/A 56 150 1.6 Motor – Planetary Gears
[48], [49]
MANUS-HAND [95] 2004 1 (P/S) N/A Ultrasonic Shinsei USR30 (~48) – 10:1 Spur Gears
1 (P/S but with
Faulhaber 0620 K 006 B (2.5) – 5:1 Belt Drive – Faulhaber
Unnamed [89] 2012 adjustable axis of N/A 40 × 40 65 87 0.0596 4.3 BLDC
08/1K 16:1 (3.8) – 10:1 Pinion and Ring Gears
rotation)
MicroMo 1724 006SR (27) – Faulhaber 415:1 gearbox –
DTM Wrist [96] 2017 1 (F/E) 90 N/A 175.47 1.0 BDC
3:1 Planetary Gearbox
Dynamixel XM-430-W210-R (105, BDCs – Gears –
Unnamed [97] 2021 1 (F/E) 90/90 N/A 43 × 29 37 125 BDC
Mounting Plate)
2 (P/S,
LUKE Arm (comm.) [10]– P/S – Motor – Harmonic Drive
Combination of Serial
[12] F/E – Motor – Gears and Non-Backdrivable Clutch
F/E and R/U)
266 – P/S,
Osaka City University Hand II
2011 3 (P/S, F/E, R/U) 175 – F/E, Serial DC Motor P/S, F/E, R/U – DC Motor – Gears
[100]
18.6 – R/U
50.4/89.3 –
Unnamed [101] 2014 2 (P/S, F/E) P/S Serial 690 DC Motor P/S, F/E – Lobot LDX-218 (60, DC Motor – Gears)
22.4/41 – F/E
45 × 35 – 58 – F/E 2.5 – F/E, 2.2 – 7.9 – F/E, F/E – BLDC Motor (26) – Planetary Gear – Non-
RIC Arm [102], [103] 2016 2 (P/S, F/E) Serial 378 BLDC
F/E Module Module P/S 8.7 – P/S backdrivable Clutch – Cycloidal Drive
2 (P/S, F/E or 180 – P/S,
Unnamed [104] 2019 Serial 74 × 74 118 330 4.3 BDC P/S, F/E – HiTEC D980TW (79, BDC – Spur Gears)
R/U) 175 – F/E
Modular Prosthetic Limb v3 P/S, F/E, R/U – BLDC Motor – 4:1 Planetary Gearbox –
2020 3 (P/S, F/E, R/U) Serial 2.1 BLDC
[13] 76:1 Cycloidal Drive
103 – P/S Pneumatic P/S – Pneumatic Cylinder – Rack and Pinion
Unnamed [105] 2014 2 (P/S, F/E) Serial 52.8 × 52.8 95.4 0.3209 – F/E
84 – F/E Cylinder F/E – Pneumatic Cylinder – Link
≥ 60 – P/S,
Unnamed [106] 2009 2 (P/S, R/U) Parallel ~0.56 DC motor P/S, R/U – DC motors – Spur Gears – Cable + Pulleys
≥ 20 – R/U
Unnamed [107], [108] 2011 2 (P/S, F/E) ≥ 40 – P/S Parallel AC motor P/S, R/U – AC Motors – Bevel Gear Differential
IEEE Transactions on Medical Robotics and Bionics 7

4.4 – P/S,
Unnamed [98] 2011 2 (P/S, F/E) 360 – P/S Parallel 60 × – 48 0.073 – F/E DC motor P/S, F/E – DC motors – Gears – Bevel Gear Differential
F/E
105 – F/E, 40
Serial and Pneumatic P/S – Bimba 021.5-DXPV – Barrel Cam
Unnamed [109] 2008 3 (P/S, F/E, R/U) – R/U, 95 –
Parallel Cylinder F/E, R/U – Bimba 022-DXPV/021-DXPV – Link
P/S
70 – P/S, P/S – Maxon RE 25 (130) – Maxon GP26B (108) – Link
Serial and
Unnamed [110] 2008 3 (P/S, F/E, R/U) 50/50 – F/E, ≤280 BDC F/E, R/U – Maxon RE 25 (130) – Maxon GP26B (108) –
Parallel
30/30 – R/U Pulleys + Cable
P/S - Faulhaber 1717012SR (18) – Faulhaber 15A 249:1
90 – F/E, 1.84 – P/S
(6) – Belt and Pulleys
Unnamed [99] 2018 3 (F/E, R/U, P/S) R/U Parallel 86 × 86 180 578 ≥ 0.6 – F/E, BDC
F/E, R/U – Actuonix P16-50-64-12-P (95, BDC – ) –
360+ – P/S R/U
Slider and Rail – Link
* 20-40% of the median forearm weight, + theoretical weights of components – measured total motor + gearbox weight = 63 g, # measured weight without base plate – designed base plate theoretically adds 11 g

rotation. In most cases, the prostheses also meet the minimum target joint speed of 2 not appear to be the case for the three 2 DoF parallel kinematic prostheses [98], [106]–
rad/s and weigh less than half of the target weight. However, 2.3 Nm was the highest [108]. Two of these prostheses actuate both P/S and F/E through two AC or DC
measured joint torque among these prostheses, substantially less than target or motors that actuate a bevel gear differential [98], [107], [108] (Fig. 2c). While this
anthropomorphic joint torques. These prostheses are also relatively long, making it leads to a simple and compact mechanical design, bevel gears typically do not provide
challenging to connect multiple of them together in series while ensuring the resulting sufficient torque density to achieve the target prosthesis specifications. Similar
prosthesis is not too long. This is likely a result of the size of the actuator and gearbox, reasons most likely limit the third prosthesis, which provides P/S and R/U (but could
which must be made more compact, torque dense (torque output per unit weight), provide F/E instead) through spur gears that in turn drive cables and pulleys [106].
power dense (power output per unit volume), and have sufficient specific power Three prostheses actuate both F/E and R/U through a PKM, with one actuating P/S
(power output per unit weight) to meet the target prosthesis specifications. This is the serially [109] in a similar manner to the human wrist and the final two actuating P/S
case for the four commercially available 1 DoF prostheses. Thus, most amputees through the same PKM [99], [110]. The first [109] actuates F/E and R/U through a
currently have to choose between having no prosthetic wrist, a flexible or manually RPR+SPS+U PKM, where the prismatic (P) joints are pneumatic cylinders that
adjustable prosthetic wrist [9], or a prosthetic wrist capable of only one (but not both) actuate the two DoFs. Actuation of P/S is provided proximally, through a third
of either F/E or P/S that provides significantly less than anthropomorphic joint torque. pneumatic cylinder that causes a barrel cam to rotate the forearm and wrist. In the
However, six prostheses utilize a motor-gearbox (including spur, planetary, second [110], F/E and R/U are provided by two BDCs and planetary gearboxes that
cycloidal, or harmonic gearboxes) combination that is compact and lightweight actuate cables and pulleys about a spherical joint. P/S is provided proximally by
enough to integrate either two or three of the same (or very similar) combinations rotating two rods with spherical joints on each end that resemble the radius and ulna.
serially into a single prosthesis [10]–[13], [100]–[104]. Motor-gearbox combinations A third BDC and planetary gearbox rotate the two rods about each other via a link to
that can achieve high torque generally are larger in diameter. For example, the produce P/S. Unlike with the first prosthesis, integrating actuation of P/S into the
combination for the RIC arm [102], [103] has a maximum width of 45mm. This same PKM means that the actuators providing F/E and R/U are not moved when
makes a SKM a logical choice for this combination as it enables each combination to moving in P/S (an example of ground-mounted actuators). This reduces the inertia
occupy the necessary space without collisions between joints; a PKM would typically required to actuate this DoF, helping to increase maximum speed and torque. The
require two combinations be able to be housed next to each other, making this final of these three prostheses [99] (Fig. 2d) is PRS+PSS+S. F/E and R/U are actuated
actuator-transmission combination impractical. As with the 1 DoF prostheses, which by two linear actuators (the two prismatic joints of the mechanism) composed of a
typically utilize similar actuator-transmission combinations, the key challenge for BDC and most likely a leadscrew and nut connecting to distal R-S and S-S links. P/S
these prosthetic wrists is maintaining both a suitable weight and torque. This is most is provided by rotating a shaft connected to the central spherical joint via a BDC
likely due to the low torque density of most gearboxes (see Section IV). A notable driving a planetary gearbox that in turn rotates the spherical joint via a belt and
exception may be the Modular Prosthetic Limb (MPL) [13] (Fig. 2b), which takes pulleys. Unlike in the two other prostheses, no part of the forearm rotates due to P/S
advantage of the relatively high torque density of cycloidal drives. However, this rotation; instead, only a prosthetic hand connected to the distal end of the prosthesis
prosthetic wrist likely does not achieve the target maximum joint torque (target of would rotate, minimizing the inertia that is actuated. Unfortunately, few
“near 8 Nm”) or prosthesis weight (prosthetic hand and wrist together weigh 1300g). specifications are provided for these prostheses. While all three are able to provide
While PKMs can be lighter weight and produce higher output torques, this does close to the functional RoM, both [99], [110] are not sufficiently compact.
IEEE Transactions on Medical Robotics and Bionics 8

Furthermore, the weight and joint torque output of [99] do not around 50 W/kg [91]), and power density of 0.025-0.19 W/cm3.
meet the targets, indicating that a more torque-dense actuator- For electromagnetic actuators such as BLDCs, which typically
transmission combination is required. Thus, while PKMs have have maximum power output near the maximum force (or
advantages over SKMs, they require actuators and torque) and maximum speed simultaneously, the target
transmissions that provide significantly higher performance to performance corresponds to a target power output of between
do so. 16-42 W, specific power 210-560 W/kg, and power density of
A key takeaway from this review is that no prosthesis is 0.23-1.7 W/cm3. In order to identify whether any existing
currently able to achieve the target specifications of the actuator-transmission combination can provide these target
prosthesis. While several can provide the target RoM and joint values, critical for producing a prosthesis with the target
speed and be sufficiently lightweight and compact, none have specifications, we first review existing actuators.
demonstrated achieving the target joint torque while
A. Review of Actuators
maintaining a sufficient weight. With the possible exception of
the MPL [13], the highest measured joint torque is 4.3 Nm [104] Pneumatic actuators, which include both pneumatic
(approximately half of the target) in a prosthesis that provides cylinders and artificial muscles (e.g. McKibben actuators), have
2 DoFs but weighs 330g (within the target weight) and is demonstrated the ability to transmit air pressure greater than 1.4
significantly larger than targeted. These findings are similar to MPa (200 psi) [112] and can achieve sufficiently high strain
the current performance limitations of prosthetic hands [8]. The rates and specific powers. While pneumatic cylinders can
current gap in performance requires an examination of actuators achieve strains greater than the target of 0.5 [112], pneumatic
and transmissions to specifically identify those that can offer artificial muscles are limited to strains of 0.25 [113] and are
sufficient torque density, power density, and specific power. therefore infeasible options. However, pneumatic actuators also
require a storage tank or compressor and valves, which when
IV. ACTUATOR AND TRANSMISSION SELECTION combined with the actuators, substantially reduce the specific
power and power density of the total system below what is
The human muscle and tendon are a uniquely impressive
necessary for the prosthesis and also introduce important safety
actuator and transmission, respectively, because they can
concerns. Precise position and velocity control are also
provide the required torque, speed, and RoM for the DoFs of
practically challenging, which also limit the dexterity of the
the wrist at a sufficiently low weight while being packed into a
prosthesis [113], [114]. These shortcomings cannot be
fraction of the overall volume of the forearm. The combination
mitigated simply with the introduction of a transmission and
of these attributes makes it very challenging to replicate the
have prevented prostheses using these actuators from achieving
human wrist’s capabilities in a prosthesis. While many artificial
the target specifications [105], [109].
actuator-transmission combinations (with requisite batteries,
Shape memory alloy (SMA) actuators are capable of
electronics, etc.) can exceed the performance of human muscles
providing sufficiently high stress, strain rate, specific power,
and tendons at larger scales (e.g. automobiles and even
and power density [91], [113], [115]–[118]. Despite providing
prosthetic legs), it is very challenging to meet them while fitting
low strains (~0.05), SMAs are small and lightweight enough to
within the limited volume of the forearm.
fold together several prosthesis lengths of the actuator (as in
An actuator and transmission with a rotary output capable of
[119]) and achieve a sufficient stroke. However, SMAs rely on
meeting the target prosthesis specifications in Table VI would
cooling to reach their original shape, and would cause a
need to provide an output of 8-12 Nm, 110° RoM, and 2-3.5
prosthesis to take several seconds to reach the nominal position
rad/s in approximately 75g (based on muscle mass for F/E and
(even with an appropriate transmission) and place additional
R/U – see Section III). Similarly, an actuator and transmission
limitations on bandwidth (i.e. limiting the ability to change
with a linear output would need to provide targets of 270-690
between wrist postures quickly) and precision [91], [113],
N, 50-86 mm, and 35-105 mm/s, depending on the target joint
[118]. SMAs are also very inefficient compared to most
torque and whether target wrist width or thickness dimensions
actuators (often less than 5% power efficiency [91], [113],
would provide the limiting constraint. These values correspond
[115]–[118]), meaning that the requisite battery and cooling
to a target specific energy of 300-460 J/kg. The target force and
components to ensure the prosthesis has ideal functionality
speed output also corresponds to a target stress of 0.85-2.2 MPa
would substantially reduce both power density and specific
assuming the maximum diameter is half the target wrist
power below the target values.
thickness and a target strain and strain rate of 0.5-0.86 and 0.35-
Piezo actuators can offer very large forces, speeds, power
1.05 lengths/s, respectively, assuming the maximum target
densities and specific powers. However, the stroke of these
prosthesis length. In comparison, human muscle can provide a
actuators is very small and therefore requires amplification
continuous maximum pressure of 0.35 MPa, strain of 0.4, and
[91]. Commercially available amplified piezos cannot achieve
strain rate of 5-10 lengths/s [91], [111].
adequate strokes while providing sufficient force, speed,
Human muscle has a maximum power output at one-third of
specific power, and power density [120]–[122]. Ultrasonic
its maximum force and one-third of its maximum speed
piezo actuators operate through different principles and thus
capability [91], [111]. The target prosthesis actuator-
can achieve sufficient displacements and speeds, but not
transmission performance would therefore correspond to a
adequate forces without a transmission [123]. However,
target power output of 1.8-4.7 W, specific power of 24-62 W/kg
electromagnetic actuators such as BDC and BLDC motors are
(continuous specific power for human muscle is typically
IEEE Transactions on Medical Robotics and Bionics 9

able to achieve higher performance in torque density, power they do offer the required capabilities as long as a transmission
density, and specific power and are therefore better options. with certain specifications can be included. BLDCs can achieve
Dielectric elastomeric actuators (DEAs) are a class of sufficient specific powers and power densities in the
particularly high-performance electroactive polymer actuators. dimensions and weight of muscles and tendons in the forearm.
DEAs function as capacitors that move based on attraction or We have found commercially available BLDCs [139], [140]
repulsion due to applied electric fields. They are typically made with values greater than 300-600 W/kg and greater than 5
of either Very High Bond (VHB) acrylic or silicone-based W/cm3, respectively. However, the large power outputs of these
elastomers and are arranged in two common architectures: motors are in a different torque and speed range than what is
spring roll [124]–[127] and stacked [128]–[130]. A key needed in a prosthesis. BLDCs that can practically fit within the
challenge that has limited the practical applications of DEAs of prosthesis and weigh less than 75 g have maximum speed
both architectures is that they have a finite maximum work outputs in the range of 8000 rad/s but maximum continuous
output. Either architecture can achieve relatively large stresses torque outputs less than 0.07 Nm [139]–[144]. These are much
or strains but not both. Thus, when spring roll DEAs were used higher speeds and lower torques than the target performance.
in an arm wrestling robot, a large number of actuators that For a BLDC to provide the target performance, transmissions
required a larger volume than the entire human arm were that can provide a significant reduction ratio (>100:1) are
needed to provide both sufficient force and stroke [125]. The therefore needed to convert this high-speed, low-torque output
maximum specific energy shown in previous articles [124], into the lower speed and higher torque of a prosthetic hand and
[125], [128]–[130] were also not sufficient to meet the target wrist. A transmission with sufficiently high torque capability at
specifications even with a high-performance transmission. a low enough weight and size can help achieve the required
Hydraulically Amplified Self-healing Electrostatic (HASEL) power density, specific power, and specific work. Therefore,
actuators bear some similarities to DEAs and are a relatively we proceed to review possible transmissions that can be used
new actuator [131]. They consist of a flexible shell filled with a with BLDC motors.
liquid dielectric and an electrode on either side. A voltage
B. Review of Transmissions
applied to the electrodes causes deformation of the shell,
leading to a strain. As with DEAs, HASEL actuators have a Spur and planetary gearboxes are a very common form of
maximum work output limitation, with an improved version transmission, used in at least 13 of the above reviewed
called Peano-HASEL actuators providing specific energy of 35 prostheses and a large number of prosthetic hands [8]. While
J/kg. This is only about 10% of the target specific energy and they can commonly be purchased already integrated with
has led to insufficient performance in a prosthetic finger [132]. BLDCs, both types of gearboxes (e.g. [139]–[142]) have poor
This performance cannot be improved with the inclusion of a torque density. For example, the lightest weight commercially
high-performance transmission. However, even though available motor-planetary gearbox combinations from Maxon
currently insufficient, these actuators do present the potential to or Faulhaber [139], [140] that could provide outputs of 8 Nm
provide significant future performance improvements through and 2 rad/s (the minimum target performance) weigh
improved manufacturing techniques that produce smaller approximately 350 g, with the motor only weighing 95 g. While
pouches (leading to higher forces) and use of materials with custom gearboxes can produce similar torque outputs for less
higher dielectric constants and breakdown voltages [131], weight, the large gear face widths needed to produce and
[133]. For example, an approximately 10x improvement in transmit such a high torque in a small diameter leads to the large
frictional shear stress was previously demonstrated in transmission weight; this factor also limits the performance of
electrostatic clutches using a similar concept [134]; however, bevel gears in a prosthetic wrist. However, planetary gearboxes
the same materials may not be feasible in Peano-HASELs. are well-suited for providing the initial reduction from the high-
Many other actuators such as magnetostrictive actuators speed, low-torque output of a BLDC to a moderate-torque (~<
[135], [136], magnetic SMAs [137], [138], and other forms of 0.2 Nm) and -speed output (~< 500 rad/s). This torque and
electroactive polymer actuators [113] have also been developed speed range capitalizes on their ability to provide a large
within the past 20 years. However, none of these actuators reduction ratio while remaining compact and lightweight. For
(along with required power electronics) have simultaneously example, the gearboxes selected for the prosthesis presented in
demonstrated sufficient force, speed, power density and this paper provide a reduction ratio of 25:1 through three stages
specific power over adequate strokes in practical applications for 26 g, providing theoretical outputs of 0.143 Nm and 264
to provide the target performance. Thus, all the actuators rad/s. This arrangement has also been used previously in several
described above will require substantial and disruptive of the above prostheses.
innovations before they can provide the specifications required Harmonic and cycloidal drives offer solutions with even
for a prosthesis. A key conclusion here is that the fundamental higher torque densities by ensuring many gear teeth are in
limitations of each of the above existing actuators cannot be contact and transmitting torque at the same time (i.e. naturally
overcome with a better transmission. Instead, limitations of high contact ratios). However, commercially available options
specific power, power density, or specific energy prevent these [145]–[149] that can provide at least 8 Nm are infeasibly heavy
actuators from achieving sufficient performance. and are often too large. While the MPL [13] presents cycloidal
Electromagnetic actuators, and more specifically BLDCs, on drives with further optimization that produce higher (but still
the other hand, are unique among artificial actuators because insufficient) torque outputs, these gearboxes are most likely too
IEEE Transactions on Medical Robotics and Bionics 10

a. b. c. capable of generating a sufficiently high force or torque from a

BLDC with a gearbox. Other commonly used options such as
U Joint
belt drives [158], lead screws [159], and chains [158], [160] do
not simultaneously provide comparable reduction ratios,
S Joints weights, and output torques to compete with gearboxes at lower
torques or with ball screws at higher torques. Unlike any other
transmission option, ball screws can therefore enable BLDCs
with gearboxes to achieve sufficiently high torque density,
P Joint
power density, specific power, and specific work in the
Fig. 3 Examples of 2 DoF Parallel Kinematic Wrists considered: a. 2-PSS+U,
b. 2-SPS+U, c. 2-PUS+U
prosthesis.

heavy to offer sufficient torque density. V. PROSTHESIS DESIGN

Worm gears are an especially popular solution in prosthetic A. Mechanism, Actuator, and Transmission Selection
hands for their non-backdrivability, high reduction ratio, and Among the commercially available BDCs and BLDCs
ability to change the axis of rotation by 90° [8]. While the worm [139]–[144], planetary gearboxes [139]–[142], and ball screws
gear is often made of steel and very compact, the wormwheel is [153]–[157] considered, the lightest weight combinations
typically made of brass or plastic to ensure the pair has capable of providing at least the minimum required force and
sufficiently high efficiency. The low strength of these materials corresponding minimum speed are Maxon BLDCs and
forces the wormwheel to have a relatively large diameter and planetary gearboxes [139] combined with KSS ball screws
face width when high torques are required. Thus, the lightest [153]. However, the dimensions of the feasible Maxon BLDCs
weight commercially available worm-wormwheel pair that can and gearboxes, which are sold integrated as a single unit, have
transmit at least 8 Nm was found to be 100 g [150]–[152]. diameters under 16mm and lengths greater than 75mm (i.e. thin
The above review of gear options demonstrates that they are and long). These dimensions make it impractical to stack
not able to achieve sufficiently high reduction ratios and torque multiple of these BLDCs, planetary gearboxes, and ball screws
densities to convert BLDC outputs into the target performance serially in an SKM while maintaining the target dimensions and
without further advances in lightweight, high-strength RoM (Table VI). Options with shorter and stockier aspect ratios
materials, gears with higher contact ratios, or gear geometries are not viable because they are substantially heavier.
and architectures. However, PKMs can incorporate ground-mounted actuators
Linkage transmissions offer a solution with a relatively high and are better-suited to integrate long and thin BLDC + gearbox
torque density by taking advantage of the small dimensions of combinations. Among the several considered 3 DoF parallel
a prosthesis. The corresponding small lengths of the links make kinematic wrists from the literature [9], [92], [161], we were
them very stiff and capable of transmitting relatively large unable to devise one that had a sufficient RoM and aspect ratio
torques. However, linkages are not able to provide a constant conducive to achieving the target specifications while
reduction ratio other than 1:1 over a sufficient RoM, preventing integrating the Maxon BLDCs and planetary gearboxes. We
them from being the transmission element that converts the therefore evaluated existing 2 DoF PKMs that could provide
output of a BLDC (potentially with a lightweight gearbox) to F/E and R/U. A future iteration of this prosthetic wrist will
the target prosthesis performance. include P/S, which is critical to a prosthetic wrist that meets
Ball screws are unique because they can transmit small input user needs (see Section IIC). P/S could be provided serially and
torques into relatively high forces with very little weight and connected proximally to the presented prosthesis (possibly via
volume. Given the significantly higher maximum efficiencies a different transmission pathway) in a similar fashion to [109],
of ball screws compared to lead screws (~90% v. 50-60% [110] and the human forearm and wrist.
[153]–[157]) for sufficiently lightweight options, we only When selecting a 2 DoF parallel kinematic wrist, several
consider ball screws in this discussion. For example, 4 mm critical requirements were considered. The mechanism needed
diameter ball screws are capable of transmitting up to 790 N to include prismatic (P) joints in order to accommodate the
[153] and will not theoretically buckle under this load or the BLDCs, planetary gearboxes, and ball screws, which have
corresponding required speed because of the relatively small linear outputs. The mechanism also needed to provide sufficient
required stroke (and therefore small ball screw length) for the load bearing and transmission capabilities, mechanical
prosthesis. Furthermore, the small length of the ball nut means advantage, RoM, and aspect ratio to achieve the target
that strains of 0.8 are feasible. These ball screws are available specifications. A mechanism with a stationary center of rotation
with leads down to 0.5 mm, which would turn 0.024 Nm into in displaced configurations was also desirable as it would be
270 N assuming 90% efficiency (~33:1 reduction ratio when more like the human wrist and therefore more intuitive for the
converted to 8 Nm). This torque is even small enough to be user to control. Among the numerous existing 2 DoF parallel
provided directly from a BLDC without any gearbox. Most kinematic wrists [9], [92], we found that PKMs consisting of a
importantly, this performance can practically be provided for single central chain providing at least two DoFs (e.g. via a
just 10g, as in the prosthesis presented in this paper. universal (U) or spherical (S) joint) and at least two actuating
At the scale and target weight of the prosthesis, we have chains that incorporate a P joint (e.g. PSS, SPS, PUS) met these
found no transmission option, other than ball screws, that is
IEEE Transactions on Medical Robotics and Bionics 11

requirements (see Fig. 3). R/U

End F/E
The 2-PSS+U mechanism (Fig. 3a) [92] was ultimately Effector
chosen because each joint offers specific functions ideal for
their locations within the mechanism. By ensuring the P joint in
the actuation chain is most proximal (in contrast to in an SPS RoM
chain), the stators of the BLDC and gearbox do not have to Limiter
move, thereby reducing the mechanism’s actuated inertia,
Universal
width, and thickness. Spherical joints were selected in the PSS
Joint
chain (e.g. as opposed to PUS or PRUU) despite engineering
challenges such as slop and limited RoM because they can S-S Link
compactly provide the load transmission capability. Since these 131 Spur Gears
chains are on the outside of the mechanism, compact spherical mm Ball Screw 107
joints can lead to a larger lever arm and therefore smaller Planetary mm
required output force from the actuator and transmission. Gearbox
However, a U joint was chosen for the central chain because it Ball Nut
can provide a larger RoM and adequate load bearing and Guide Rail
transmission capabilities despite being larger. Since the central
Base Plate
chain is responsible for providing the constraints within the
mechanism [92], the lower slop of the U joint is an additional BLDC Motor
advantage. The U joint also has axes of rotation that intersect, Encoder
ensuring the center of rotation of the mechanism is stationary in 66 mm 52 mm
all displaced configurations. Fig. 4 Prosthesis Design: Top view (left), Side view (right)
The chosen mechanism is also very similar to the human
wrist. The wrist joint is kinematically modeled as a universal [153]), to an output force and speed Fb and vb of:
joint while the four muscles that actuate it are located on the 2𝜋𝜏𝑔 𝜂𝑏 𝜔𝑔 ℓ𝑏
𝐹𝑏 = = 360 N, 𝑣𝑏 = = 85 mm/s
outside of the forearm and are functionally similarly to actuated ℓ𝑏 2𝜋
prismatic joints. The tendons that serve as the transmission Assuming a lever arm of rw = 27 mm, 45% of the maximum
connecting the muscles and wrist joint are analogous to cables target wrist width (to account for the width of a S-S link), this
and have a similar function to the spherical joints (together corresponds to prosthesis maximum continuous output torque
called an S-S link). The key difference between them is that and speed τw and ωw of:
𝑣𝑏
tendons are only semi-rigid in tension while the link connecting 𝜏𝑤 = 𝑟𝑤 × 𝐹𝑏 = 9.7 Nm, 𝜔𝑤 = = 3.1 𝑟𝑎𝑑/𝑠
the spherical joints in the S-S link is rigid in tension, 𝑟𝑤
compression, and bending. This reduces the number of in the nominal position; both are within target performance.
actuators from four muscles for a human to two BLDCs for the B. Detailed Design Overview
prosthesis and therefore also reduces the required power
Fig. 4 illustrates an overview of the prosthesis design. The
density, torque density, and specific power. However, the
prosthesis from the top view (left) is symmetric about the
ability of tendons to bend (i.e. low bending stiffness) helps to
illustrated dashed black vertical center line except for the
increase the RoM of the human wrist and maintains a higher
universal joint. It is therefore possible to rotate the wrist 180°
mechanical advantage in displaced configurations (e.g. fully
about this line, swapping Flexion with Extension and likewise
flexed) compared to the 2-PSS+U mechanism. An S-S link with
Radial and Ulnar Deviation. The prosthesis is oriented such that
finite bending stiffness could therefore offer potential
the End Effector is the most distal component from a user’s
performance improvements that will be explored in future
elbow (see Fig. 5). The End Effector could include an integrated
research.
interface (e.g. [162]) that would connect to a prosthetic hand.
After selection of the 2-PSS+U mechanism, the Maxon ECX
The space around the universal joint is also available and could
Speed 13L HP BLDC, Maxon GPX 13 Speed 25:1 Planetary
be used to allow wires necessary to power and control the
Gearbox, and KSS SR0402-C7 ball screw and nut were found
prosthetic hand to pass through easily without sharp bends.
to be the lightest weight actuator and transmission combination
Finally, the prosthetic wrist could be secured to the user through
capable of providing the target speed and torque. The BLDC’s
a socket or connect to more proximal prosthesis components
listed maximum continuous torque is τm = 0.00715 Nm and
such as a P/S prosthetic wrist or prosthetic elbow via a rigid
maximum continuous speed is ωm = 6600 rad/s. The planetary
connection to the base plate.
gearbox has reduction ratio Ng = 25 and theoretical efficiency
Actuation in both DoFs is provided by the two BLDC motors
ηg of 80% (reasonable for Maxon 2-stage planetary gearboxes
(as seen in Fig. 4), each connected to a Planetary Gearbox. The
[139]), providing an output torque and speed τg and ωg of:
𝜔𝑚 output shaft of each gearbox is connected to a spur gear via a
𝜏𝑔 = 𝜏𝑚 𝜂𝑔 𝑁𝑔 = 0.143 𝑁𝑚, 𝜔𝑔 = = 264 𝑟𝑎𝑑/𝑠 shaft coupler. Each spur gear drives a second spur gear that is
𝑁𝑔
in turn coupled to a ball screw. Rotation of each ball screw
The gearbox outputs are converted by the ball screw, with lead drives the linear motion of a ball nut, which is constrained to
ℓb = 2 mm and ηb of 80% (conservative for KSS ball screws
IEEE Transactions on Medical Robotics and Bionics 12

translate by a guide rail. Each ball nut is also connected to an

S-S Link; the S-S Links are also connected to the End Effector,
which is constrained to rotate in F/E and R/U by the universal
joint in the center (and forming the 2-PSS+U mechanism).
Thus, F/E is provided by driving the BLDCs such that both ball
nuts translate in the same direction (both up or both down in
Fig. 4). R/U is provided when the ball nuts are driven to
translate in opposite directions (one up and one down in Fig. 4).
A significant consequence of the actuator and transmission
selection is that the total length of the BLDC, planetary gearbox
(including output shaft), and ball screw is 168 mm. Given the
maximum target length is only 100 mm, the ball screw must be
kept parallel and next to the BLDC and gearbox instead of being
collinear (as would be possible with a shorter BLDC and
gearbox). We therefore selected lightweight pairs of brass spur
gears (SDP/SI A 1B 2MYK08020), modified to reduce the face
widths and remove the hub. The resultant mass of each pair of Fig. 5 Prosthesis without wires and base plate overlaid on an approximately
spur gears was 8 g (4 g per gear), which is inconsequential to median male wrist and forearm extending to the elbow on the right
the actuator and transmission selection.
However, not keeping the BLDC and gearbox collinear to the large as the RoM of the central Universal Joint, the RoM of the
ball screw leads to the prosthesis’ larger than target dimensions. selected spherical joints (59935K14 – McMaster-Carr, USA) is
Spatial constraints in both width and thickness caused by the a relatively small swivel angle of 30°, the highest found among
BLDCs being placed parallel to the ball screws and S-S links sufficiently compact spherical joints. RoM of the prosthesis is
lead to the prosthesis being wider and thicker than targeted (66 therefore currently limited by these components, which is
× 52 mm rather than 55-60 × 35-40 mm). Furthermore, the enforced by the RoM Limiter (i.e., mechanical hard-stop). In the
BLDC and gearbox protrude further than any other component future, custom-made spherical joints with larger RoM (e.g.
on the proximal side of the prosthesis and therefore dictate the [163], [164]) could be used to ensure the prosthesis RoM meets
prosthesis’ length (131 mm rather than 70-100 mm). However, the target values. The selected Universal Joint, the RS Pro
the practical length of the prosthesis is shorter than the 7906699 Universal Joint, is the lightest weight option that was
measured length. The center of the universal joint is the location found with ample load rating and RoM. It has a torque
of intersection of the DoFs of the wrist [92] and the prosthesis transmission rating of 600 Nm, which theoretically translates to
would be positioned such that this location coincides with the an axial load rating over 1 kN.
virtual center of the human wrist (as shown in Fig. 5). Thus, the Fabrication of the prosthesis was primarily carried out using
End Effector and distal half of the universal joint could be traditional fabrication techniques on a mill and lathe. Several
integrated into the prosthetic hand and the prosthetic wrist additional components including the End Effector, components
would only need a length of 107 mm in the forearm. mounting the motors and ball screws, RoM Limiter, and shaft
The aluminum shaft guide rails run parallel to each ball couplers were 3D-printed using the Formlabs Form 3 and UV-
screw to constrain the ball nut to translate and to absorb any cured Rigid 4000 Resin, a high-strength plastic resin. The ball
potential radial load, which the ball screw and nut are not rated screws were cut to length and modified to have custom journal
for. An Oilite bushing serves as the linear bearing connecting ends on a manual surface grinder because they are made of a
the guide rail, ball nut, and S-S link via aluminum plates. The hardened carbon steel (SCM415) that could not be machined on
ball nut is only constrained to these components along the ball a lathe with carbide insert tooling.
screw direction of travel (vertically in Fig. 4). Undersized
shoulder bolts secured to the aluminum plates ensure the ball VI. PROSTHETIC WRIST EVALUATION
nut can float in the other two directions (right/left and into the The prosthesis is shown without its wires and base plate
page in Fig. 4) and isolate it from radial loads. This approach overlaid on an approximately median male wrist and forearm in
for a guide rail is different from that used in [99], which uses Fig. 5 to demonstrate its relative size. The fabricated prosthesis
commercially available precision guide rails and sliders. While includes an integrated prosthetic gripper not presented in this
precision guide rails and sliders are more efficient, they are paper (see Fig. 6), which shares a common, larger base plate
significantly larger and would prevent the prosthetic wrist from than that illustrated in Fig. 4. The measured prosthesis weight
meeting target dimensions (as in [99]). More compact, circular of 320 g listed in Table VII therefore does not include this
linear bearings such as those used in many 3D printers were also weight, which would theoretically lead the prosthesis to weigh
considered but options that were sufficiently compact were not 331 g. Given the presence of this prosthetic gripper, the
found because of the spatial constraints caused by the non- prosthesis was operated with the End Effector mounted into a
collinear BLDC, gearbox, and ball screw. vice and the rest of the prosthesis moving (i.e. backwards to
Spherical joint selection was primarily dictated by RoM. how it is presented in Fig. 5, as shown in Fig. 6). The total
While the required RoM for the joints of the S-S link are not as moving weight during evaluation is therefore approximately
IEEE Transactions on Medical Robotics and Bionics 13

600 g, a representative weight for if the prosthetic wrist was Fishing Line
oriented as intended and articulating an average prosthetic hand
Load Cell
(that would be attached to the End Effector) holding a Prosthetic
lightweight 100-200 g object. Gripper
Vices
RoM was measured via still images of the prosthesis
positioned in the four extreme angles using Adobe Photoshop
2021. The differences in RoM between Flexion and Extension Prosthetic
(48°/33° in F/E) and Radial and Ulnar Deviation (33°/25° in Wrist
R/U) can be attributed to the functionality of a universal joint.
The first of the joint’s two rotations is about a fixed axis relative
to ground while the second is about an axis that rotates about
the first axis. This places different limits on the RoM of the S-
S links’ joints in Flexion versus Extension and similarly for
R/U. The axes of the universal joint were therefore positioned
Fig. 6 Evaluation Setup for Maximum Torque Output Measurement
to maximize Flexion, which has the largest Functional RoM.
The prosthesis surpasses the target Radial Deviation (25°) and
a measured force of under 20 g. A current impulse was
is within 15% of the target Flexion (55°). However, further
commanded and held for at least 2 s and the maximum
improvements to the RoM of the S-S link spherical joints (as
measured force was recorded. When the commanded current
has been demonstrated previously [163], [164] ) will be needed
impulse was the continuous rated current of 2.2 A for the
to achieve the target RoM in both directions of each DoF.
BLDCs, the torque was measured to be 7.0 Nm in R/U and 7.5
For speed and torque evaluation, the BLDCs were each
Nm in F/E. However, the BLDCs were not appreciably hotter
individually current driven by an Advanced Motion Controls
than previously, demonstrating that the maximum torque output
AZB10A4 PWM servo drive and MC1XAZ02 mounting card,
is higher for at least a period of 2s. Current impulses up to the
each individually powered by a MeanWell LRS-75-24 24V DC
power supply current rating of 3.2 A were provided, leading to
power supply and controlled simultaneously by a single NI
the measurements of 8.4 Nm in R/U and 8.2 Nm in F/E listed
myRIO operating NI LabVIEW. Position feedback of each ball
in Table VII. This is therefore the first prosthetic wrist to exceed
screw is provided by 12 CPR magnetic encoders (Pololu 3081),
the minimum target joint torque of 8 Nm. However, the BLDCs
with encoder wheels press fit onto one end of each ball screw.
were still not appreciably hotter after these current impulses and
Maximum speed in F/E and R/U without any additional loads
even higher torques could possibly be achieved. This could be
applied to the prosthesis beyond the 600g moving mass
applicable in brief tasks the user may perform when larger
(representative of a prosthetic hand holding a small weight)
torques are required (e.g. short-term heavy lifting).
were measured individually from video footage recorded on an
The substantially lower measured torque values at the rated
Apple iPhone 10 at 30 fps (see Supplementary Video 1). For
continuous current of the BLDCs (compared to the theoretical
measurements in F/E, a current impulse was sent to move the
value of 9.7 Nm) indicates that the mechanism is not as efficient
prosthesis from fully flexed to fully extended and vice versa;
as initially designed. While the efficiency of the planetary
analogous measurements were conducted in R/U. The change
gearbox or ball screws and nuts could be lower than predicted,
in angle between frames was measured using Adobe Premiere
it is likely that this reduced efficiency is from other sources.
Pro 2022 to calculate the maximum speed. The prosthesis
These sources could include the spur gears but is more likely to
achieved a measured maximum speed of 3.1 rad/s at the
be from the Oilite bushings used as the linear bearings for the
nominal configuration and 4.2 rad/s at displaced configurations,
guide rails. These bushings will be replaced with more efficient
meeting or exceeding the target speed range of 2-3.5 rad/s. In
bearings in future versions of this prosthesis.
order to evaluate the prosthesis’ performance in both DoFs, the
The combined weight of the actuators and transmissions
prosthesis was commanded to trace a cone at a set frequency.
excluding the S-S links (which are analogous to tendons in
The prosthesis was able to trace a cone with a 23° angle from
function) is 162 g. This is comparable to the weight of the
vertical at a frequency of 1.5 Hz (see Supplementary Video 1).
muscles that actuate F/E and R/U in the human wrist (153 g ±
Maximum torque output was measured while the prosthesis
69 g). While the prosthesis cannot achieve True Maximum male
was not moving (i.e. stall torque while articulating 600g) by
joint torques, a heavier and larger prosthetic wrist (feasible for
securing a Yo-Zuri SuperBraid 50 lb. Braided Fishing Line tied
a median male) with larger BLDCs and planetary gearboxes
to the prosthesis on one end and on the other to a TAL220 10
could supply adequately high torque outputs. However, despite
kg straight bar load cell bolted into a rigid plank secured in a
the high torque and speed performance of this prosthesis, which
vice (Fig. 6). The load cell measurements were recorded on an
is the first to meet both target values, this actuator and
Arduino Uno from a HX711 Sparkfun Load Cell Amplifier
transmission combination does not match the human wrist in
connected to the load cell. Before measurements were taken,
maximum speed. While human wrist speeds are currently too
the load cell was calibrated using standard weights. Lever arms
fast for a prosthetic hand user to control (see Section IIC) and
were measured using digital calipers. Before each joint torque
thus not necessary in this prosthesis, if matching human muscle
measurement was completed, the prosthesis was positioned in
capabilities at the scale of forearm muscles were desirable, a
the nominal position, the cable was slack, and the load cell had
IEEE Transactions on Medical Robotics and Bionics 14

substantial improvement in power density and specific power ACKNOWLEDGEMENTS

would be required in order to match this performance. We would like to thank Megan Diemer and Alicia Davis for
The weight of the actuators and transmissions means that their contributions in providing feedback on the prosthesis. We
approximately half of the prosthesis’ weight is used for other would also like to thank Kent Pruss, Charlie Bradley, and
purposes such as critical mechanical components (e.g. the guide Jonathon Yenkel for their contributions in fabricating the
rail, base plate, and universal and spherical joints). This is prosthesis prototype.
relatively high for a prosthesis and likely high relative to the
human forearm and wrist. There is therefore potential for it to REFERENCES
be substantially reduced through design changes including [1] C. W. Heckathorne, “Upper-Limb Prosthetics: Components for Adult
more custom rather than off-the-shelf components (e.g. the Externally Powered Systems,” in Atlas of Limb Prosthetics: Surgical,
universal and spherical joints) rated specifically for the target Prosthetic, and Rehabilitation Principles, 2nd ed., Rosemont, IL: Academy
specifications, better material selection (e.g. high performance of Orthopedic Surgeons, 1992.
[2] R. F. Weir, “The Design of Artificial Arms and Hands for Prosthetic
plastics or composites with higher strength-to-density ratios Applications,” in Standard Handbook of Biomedical Engineering and
than metals), and improved optimization. Design, M. Kutz, Ed. New York: McGraw-Hill, 2003, pp. 32.1-32.59.
[3] J. T. Belter, J. L. Segil, A. M. Dollar, and R. F. Weir, “Mechanical design
and performance specifications of anthropomorphic prosthetic hands: A
VII. CONCLUSION AND FUTURE WORK review,” J. Rehabil. Res. Dev., vol. 50, no. 5, pp. 599–618, 2013, doi:
The prosthesis presented in this paper is the first to achieve 10.1682/JRRD.2011.10.0188.
[4] C. Piazza, G. Grioli, M. G. Catalano, and A. Bicchi, “A Century of Robotic
target performance in both speed and torque while achieving Hands,” Annu. Rev. Control. Robot. Auton. Syst., vol. 2, no. 1, pp. 1–32,
the target weight. It is therefore an important first step toward May 2019, doi: 10.1146/annurev-control-060117-105003.
developing a prosthetic wrist that will ultimately meet user [5] I. Vujaklija, D. Farina, and O. C. Aszmann, “New developments in
prosthetic arm systems,” Orthopedic Research and Reviews, vol. 8. Dove
needs and be adopted by prosthetic hand users. Its performance Medical Press Ltd., pp. 31–39, 07-Jul-2016, doi: 10.2147/ORR.S71468.
is the result of using actuators and transmissions that have [6] J. ten Kate, G. Smit, and P. Breedveld, “3D-printed upper limb prostheses:
sufficient power density, torque density, and specific power as a review,” Disability and Rehabilitation: Assistive Technology, vol. 12, no.
3. Taylor and Francis Ltd, pp. 300–314, 03-Apr-2017, doi:
well as a mechanism that offers sufficient load bearing and 10.1080/17483107.2016.1253117.
transmission capability in a relatively small weight and close to [7] V. Mendez, F. Iberite, S. Shokur, and S. Micera, “Current Solutions and
sufficiently compact shape. Future Trends for Robotic Prosthetic Hands,” Annu. Rev. Control. Robot.
Auton. Syst., vol. 4, no. 1, pp. 595–627, 2021, doi: 10.1146/annurev-
However, this prosthesis does not meet target dimensions control-071020-104336.
because of the shape of commercially available BLDCs and [8] R. Damerla, Y. Qiu, T. M. Sun, and S. Awtar, “A Review of the
planetary gearboxes, which are smaller in diameter and larger Performance of Extrinsically Powered Prosthetic Hands,” IEEE Trans.
in length than would be ideal. We are currently investigating Med. Robot. Bionics, vol. 3, no. 3, pp. 640–660, Jul. 2021, doi:
10.1109/TMRB.2021.3100612.
how to utilize BLDCs that are larger in diameter and smaller in [9] N. M. Bajaj, A. J. Spiers, and A. M. Dollar, “State of the Art in Artificial
length that would enable a collinear BLDC, gearbox, and ball Wrists: A Review of Prosthetic and Robotic Wrist Design,” IEEE Trans.
screw. These BLDCs would also naturally be more torque Robot., vol. 35, no. 1, pp. 261–277, Feb. 2019, doi:
10.1109/TRO.2018.2865890.
dense (although potentially at the cost of lower specific power), [10] “Mobius Bionics LUKE Arm Product Specification Sheet,” Mobius
reducing the size, weight, and number of stages for the Bionics, Manchester, NH, USA. [Online]. Available:
planetary gearbox and thereby increasing its efficiency. Options https://www.mobiusbionics.com/luke-arm/
[11] L. Resnik, S. L. Klinger, and K. Etter, “The DEKA Arm: Its features,
such as drone motors have been utilized in prosthetic legs [165] functionality, and evolution during the veterans affairs study to optimize
and may offer promise in prosthetic wrists as well. If a BLDC the DEKA Arm,” Prosthet. Orthot. Int., vol. 38, no. 6, pp. 492–504, Dec.
and planetary gearbox could be sufficiently shorter in length to 2014, doi: 10.1177/0309364613506913.
[12] C. O. Evans et al., “Arm Prosthetic Device,” U.S. Patent 8,979,943 B2,
be collinear to the ball screw, it would lead to a substantially 2015.
more compact prosthesis. RoM is similarly limited by the RoM [13] M. S. Johannes et al., “The modular prosthetic limb,” in Wearable
of commercially available spherical joints, which can be Robotics: Systems and Applications, Elsevier, 2020, pp. 393–444.
[14] K. Ziegler-Graham, E. J. MacKenzie, P. L. Ephraim, T. G. Travison, and
improved through the integration of smaller custom joints. R. Brookmeyer, “Estimating the Prevalence of Limb Loss in the United
In its current form, this prosthesis also lacks the requisite States: 2005 to 2050,” Arch. Phys. Med. Rehabil., vol. 89, no. 3, pp. 422–
integrated motor drivers and controllers needed to be used in 429, Mar. 2008, doi: 10.1016/j.apmr.2007.11.005.
any practical scenario. However, multiple prostheses have [15] F. Cordella et al., “Literature Review on Needs of Upper Limb Prosthesis
Users,” Front. Neurosci., vol. 0, no. MAY, p. 209, 2016, doi:
demonstrated the feasibility of integrating custom electronic 10.3389/FNINS.2016.00209.
components into a prosthesis for minimal additional size or [16] I. Jeon et al., “Trends in the incidence of work-related traumatic limb
weight [13], [102], [103]. Finally, it is also critical to include amputations in South Korea from 2004 to 2013:,” Prosthet. Orthot. Int.,
vol. 43, no. 4, pp. 409–417, Mar. 2019, doi: 10.1177/0309364619838356.
P/S in any prosthetic wrist. While the dimensional constraints [17] M. Moini, M. R. Rasouli, A. Khaji, F. Farshidfar, and P. Heidari, “Patterns
available for P/S are very different than for the prosthesis of extremity traumas leading to amputation in Iran: results of Iranian
presented in this paper, the same actuator and transmission National Trauma Project,” Chinese J. Traumatol. (English Ed.), vol. 12,
no. 2, pp. 77–80, Apr. 2009, doi: 10.3760/CMA.J.ISSN.1008-
considerations used to achieve both sufficient speed and torque 1275.2009.02.003.
should be reflected in the mechanism selection and overall [18] L. Resnik, J. Cancio, S. Klinger, G. Latlief, N. Sasson, and L. Smurr-
design. Without these considerations, it is unlikely that P/S Walters, “Predictors of retention and attrition in a study of an advanced
upper limb prosthesis: implications for adoption of the DEKA Arm,”
could be achieved with sufficient specifications. Disabil. Rehabil. Assist. Technol., vol. 13, no. 2, pp. 206–210, Feb. 2018,
IEEE Transactions on Medical Robotics and Bionics 15

doi: 10.1080/17483107.2017.1304585. 10.1016/j.clinbiomech.2008.05.008.

[19] L. V. Mcfarland, S. L. H. Winkler, A. W. Heinemann, M. Jones, and A. [39] F. Montagnani, M. Controzzi, and C. Cipriani, “Is it Finger or Wrist
Esquenazi, “Unilateral upper-limb loss: Satisfaction and prosthetic- Dexterity That is Missing in Current Hand Prostheses?,” IEEE Trans.
device use in veterans and servicemembers from Vietnam and OIF/OEF Neural Syst. Rehabil. Eng., vol. 23, no. 4, pp. 600–609, 2015, doi:
conflicts,” J. Rehabil. Res. Dev., vol. 47, no. 4, pp. 299–316, 2010, doi: 10.1109/TNSRE.2015.2398112.
10.1682/JRRD.2009.03.0027. [40] A. J. Spiers, Y. Gloumakov, and A. M. Dollar, “Examining the Impact of
[20] E. Biddiss and T. Chau, “Upper limb prosthesis use and abandonment: A Wrist Mobility on Reaching Motion Compensation Across a Discretely
survey of the last 25 years,” Prosthet. Orthot. Int., vol. 31, no. 3. SAGE Sampled Workspace,” in Proceedings of the IEEE RAS and EMBS
PublicationsSage UK: London, England, pp. 236–257, 23-Sep-2007, doi: International Conference on Biomedical Robotics and Biomechatronics,
10.1080/03093640600994581. Biorob 2018, 2018, pp. 819–826, doi: 10.1109/BIOROB.2018.8487871.
[21] L. Resnik, S. Ekerholm, M. Borgia, and M. A. Clark, “A national study [41] M. Deijs, R. M. Bongers, N. D. M. Ringeling-Van Leusen, and C. K. Van
of Veterans with major upper limb amputation: Survey methods, Der Sluis, “Flexible and static wrist units in upper limb prosthesis users:
participants, and summary findings,” PLoS One, vol. 14, no. 3, Mar. 2019, Functionality scores, user satisfaction and compensatory movements,” J.
doi: 10.1371/journal.pone.0213578. Neuroeng. Rehabil., vol. 13, pp. 1–13, 2016, doi: 10.1186/s12984-016-
[22] E. Biddiss, D. Beaton, and T. Chau, “Consumer design priorities for upper 0130-0.
limb prosthetics,” Disabil. Rehabil. Assist. Technol., vol. 2, no. 6, pp. [42] G. Kanitz, F. Montagnani, M. Controzzi, and C. Cipriani, “Compliant
346–357, 2007, doi: 10.1080/17483100701714733. Prosthetic Wrists Entail More Natural Use Than Stiff Wrists during
[23] A. Chadwell et al., “Upper limb activity in myoelectric prosthesis users is Reaching, Not (Necessarily) during Manipulation,” IEEE Trans. Neural
biased towards the intact limb and appears unrelated to goal-directed task Syst. Rehabil. Eng., vol. 26, no. 7, pp. 1407–1413, 2018, doi:
performance,” Sci. Reports, vol. 8, no. 1, pp. 1–12, Jul. 2018, doi: 10.1109/TNSRE.2018.2847565.
10.1038/s41598-018-29503-6. [43] P. J. Kyberd, “The influence of passive wrist joints on the functionality of
[24] A. J. Spiers, J. Cochran, L. Resnik, and A. M. Dollar, “Quantifying prosthetic hands,” Prosthet. Orthot. Int., vol. 36, no. 1, pp. 33–38, 2012,
Prosthetic and Intact Limb Use in Upper Limb Amputees via Egocentric doi: 10.1177/0309364611426905.
Video: An Unsupervised, At-Home Study,” IEEE Trans. Med. Robot. [44] T. Bertels, T. Schmalz, and E. Ludwigs, “Objectifying the Functional
Bionics, vol. 3, no. 2, pp. 463–484, 2021, doi: Advantages of Prosthetic Wrist Flexion,” J. Prosthetics Orthot., vol. 21,
10.1109/TMRB.2021.3072253. no. 2, pp. 74–78, Apr. 2009, doi: 10.1097/JPO.0b013e3181a10f46.
[25] L. E. Jones and J. H. Davidson, “Save that arm: a study,” Prosthet. Orthot. [45] H. H. Sears, E. Iversen, S. Archer, and T. Jacobs, “Wrist Innovations To
Int., vol. 23, pp. 55–58, 1999. Improve Function of Electric Terminal Devices,” in Proceedings of the
[26] D. Datta, K. Selvarajah, and N. Davey, “Functional outcome of patients 2008 MyoElectric Controls/Powered Prosthetics Symposium, 2008, pp.
with proximal upper limb deficiency - Acquired and congenital,” Clin. 1–4.
Rehabil., vol. 18, no. 2, pp. 172–177, 2004, doi: [46] “MC Standard and ProWrist Rotators Fact Sheet,” Fillauer, Chattanooga,
10.1191/0269215504cr716oa. TN, USA. [Online]. Available: https://fillauer.com/wp-
[27] C. R. Gambrell, “Overuse Syndrome and the Unilateral Upper Limb content/uploads/2020/08/1910062-FACT-SHEET-Wrist-Rotator-Rev-E-
Amputee: Consequences and Prevention,” J. Prosthetics Orthot., vol. 20, 08-26-2020.pdf
no. 3, pp. 126–132, Jul. 2008, doi: 10.1097/JPO.0b013e31817ecb16. [47] “Ottobock Electric Wrist Rotator Flyer,” Ottobock, Duderstadt, DEU.
[28] K. Østlie, R. J. Franklin, O. H. Skjeldal, A. Skrondal, and P. Magnus, [Online]. Available: https://shop.ottobock.us/Prosthetics/Upper-Limb-
“Musculoskeletal pain and overuse syndromes in adult acquired major Prosthetics/Myo-Hands-and-Components/Myo-Wrist-Units-and-
upper-limb amputees,” Arch. Phys. Med. Rehabil., vol. 92, no. 12, pp. Rotation/Electric-Wrist-Rotator/p/10S17#product-documents-section
1967-1973.e1, 2011, doi: 10.1016/j.apmr.2011.06.026. [48] “i-Limb Wrist Technical Manual,” Ӧssur, Reykjavik, ISL. [Online].
[29] L. C. Smail, C. Neal, C. Wilkins, and T. L. Packham, “Comfort and Available: https://training.touchbionics.com/pdfs/1421_002_i-
function remain key factors in upper limb prosthetic abandonment: Limb_Wrist_Technical_Manual.pdf
findings of a scoping review,” Disabil. Rehabil. Assist. Technol., 2020, [49] H. Gill, “Systems and Methods for Prosthetic Wrist Rotation,” U.S. Patent
doi: 10.1080/17483107.2020.1738567. 10,369,024 B2, 2019.
[30] S. M. Engdahl, B. P. Christie, B. Kelly, A. Davis, C. A. Chestek, and D. [50] “Motion Control Powered Flexion Wrist Flyer,” Fillauer, Chattanooga,
H. Gates, “Surveying the interest of individuals with upper limb loss in TN, USA. [Online]. Available: https://fillauer.com/products/mc-
novel prosthetic control techniques,” J. Neuroeng. Rehabil., vol. 12, no. powered-flexion-wrist/
1, Jun. 2015, doi: 10.1186/s12984-015-0044-2. [51] E. K. Iversen, J. Christensen, and G. J. Jacobs, “Powered Prosthetic
[31] C. Pylatiuk, S. Schulz, and L. Döderlein, “Results of an internet survey of Flexion Device,” U.S. Patent 2020/0375761 A1, 2020.
myoelectric prosthetic hand users,” Prosthet. Orthot. Int., vol. 31, no. 4, [52] B. Hirt, H. Seyhan, M. Wagner, and R. Zumbasch, “Wrist and Movement
pp. 362–370, Dec. 2007, doi: 10.1080/03093640601061265. Axes,” in Hand and Wrist Anatomy and Biomechanics: A Comprehensive
[32] L. Resnik, H. Benz, M. Borgia, and M. A. Clark, “Patient perspectives on Guide, 1st ed., Stuttgart, Germany, 2017, pp. 10–47.
benefits and risks of implantable interfaces for upper limb prostheses: a [53] B. Hirt, H. Seyhan, M. Wagner, and R. Zumbasch, “Anatomy and
national survey,” Expert Rev. Med. Devices, vol. 16, no. 6, 2019, doi: Functional Anatomy of the Hand,” in Hand and Wrist Anatomy and
10.1080/17434440.2019.1619453. Biomechanics: A Comprehensive Guide, 1st ed., Stuttgart, Germany,
[33] M. Espinosa and D. Nathan-Roberts, “Understanding Prosthetic 2017, pp. 2–9.
Abandonment,” Proc. Hum. Factors Ergon. Soc. Annu. Meet., vol. 63, no. [54] B. Hirt, H. Seyhan, M. Wagner, and R. Zumbasch, “Structure and
1, pp. 1644–1648, Nov. 2019, doi: 10.1177/1071181319631508. Function of the Finger Joints,” in Hand and Wrist Anatomy and
[34] B. Peerdeman et al., “Myoelectric forearm prostheses: State of the art Biomechanics: A Comprehensive Guide, 1st ed., Stuttgart, Germany:
from a user-centered perspective,” J. Rehabil. Res. Dev., vol. 48, no. 6, p. Thieme, 2017, pp. 66–80.
719, Jan. 2011, doi: 10.1682/JRRD.2010.08.0161. [55] J. W. Garrett, “Anthropometry of the Air Force Female Hand,” Mar. 1970.
[35] A. J. Metzger, A. W. Dromerick, R. J. Holley, and P. S. Lum, [56] J. W. Garrett, “Anthropometry of the Hands of Male Air Force Flight
“Characterization of compensatory trunk movements during prosthetic Personnel,” 1970.
upper limb reaching tasks,” Arch. Phys. Med. Rehabil., vol. 93, no. 11, [57] R. J. Marley and M. R. Thomson, “Isokinetic Strength Characteristics in
pp. 2029–2034, 2012, doi: 10.1016/j.apmr.2012.03.011. Wrist Flexion and Extension,” Int. J. Ind. Ergon., vol. 25, no. 6, pp. 633–
[36] M. J. Major, R. L. Stine, C. W. Heckathorne, S. Fatone, and S. A. Gard, 643, 2000.
“Comparison of range-of-motion and variability in upper body [58] T. Xia and L. A. Frey-Law, “Wrist joint torque-angle-velocity
movements between transradial prosthesis users and able-bodied controls performance capacity envelope evaluation and modelling,” Int. J. Hum.
when executing goal-oriented tasks,” J. Neuroeng. Rehabil., vol. 11, pp. Factors Model. Simul., vol. 5, no. 1, pp. 33–52, 2015.
1–10, 2014, doi: 10.1186/1743-0003-11-132. [59] T. M. Greiner, “Hand Anthropometry of U.S. Army Personnel,” 1991.
[37] A. Hussaini, A. Zinck, and P. Kyberd, “Categorization of compensatory [60] R. M. White, “Comparative Anthropometry Of The Hand,” 1981.
motions in transradial myoelectric prosthesis users,” Prosthet. Orthot. [61] C. C. Gordon et al., “2012 Anthropometric Survey of U.S. Army
Int., vol. 41, no. 3, pp. 286–293, 2017, doi: 10.1177/0309364616660248. Personnel: Methods and Summary Statistics,” 2014.
[38] S. L. Carey, M. Jason Highsmith, M. E. Maitland, and R. V. Dubey, [62] R. W. Schoenmarklin and W. S. Marras, “Dynamic capabilities of the
“Compensatory movements of transradial prosthesis users during wrist joint in industrial workers,” Int. J. Ind. Ergon., vol. 11, no. 3, pp.
common tasks,” Clin. Biomech., vol. 23, no. 9, pp. 1128–1135, 2008, doi: 207–224, 1993, doi: 10.1016/0169-8141(93)90109-Q.
IEEE Transactions on Medical Robotics and Bionics 16

[63] J. T. Barter, I. Emanuel, and B. Truett, “A statistical evaluation of joint of wrist flexion and extension torques in different forearm positions,”
range data,” 1957. Biomed. Eng. Online, vol. 14, pp. 1–10, 2015, doi: 10.1186/s12938-015-
[64] A. R. Bonebrake, J. E. Fernandez, R. J. Marley, J. B. Dahalan, and K. J. 0110-9.
Kilmer, “A treatment for carpal tunnel syndrome: Evaluation of objective [88] V. Decostre et al., “Wrist flexion and extension torques measured by
and subjective measures,” J. Manipulative Physiol. Ther., vol. 13, no. 9, highly sensitive dynamometer in healthy subjects from 5 to 80 years,”
pp. 507–520, 1990. BMC Musculoskelet. Disord., vol. 16, pp. 1–10, 2015, doi:
[65] J. M. Soucie et al., “Range of motion measurements: Reference values 10.1186/s12891-015-0458-9.
and a database for comparison studies,” Haemophilia, vol. 17, no. 3, pp. [89] A. Zinck, Ø. Stavdahl, E. Biden, and P. J. Kyberd, “Design of a compact,
500–507, 2011, doi: 10.1111/j.1365-2516.2010.02399.x. reconfigurable, prosthetic wrist,” Appl. Bionics Biomech., vol. 9, pp. 117–
[66] M. M. Marshall, J. R. Mozrall, and J. E. Shealy, “The effects of complex 124, 2012, doi: 10.3233/ABB-2011-0043.
wrist and forearm posture on wrist range of motion,” Hum. Factors, vol. [90] K. R. S. Holzbaur, W. M. Murray, G. E. Gold, and S. L. Delp, “Upper
41, no. 2, pp. 205–213, 1999, doi: 10.1518/001872099779591178. limb muscle volumes in adult subjects,” J. Biomech., vol. 40, no. 4. pp.
[67] D. C. Boone and S. P. Azen, “Normal range of motion of joints in male 742–749, 2007, doi: 10.1016/j.jbiomech.2006.11.011.
subjects,” J. Bone Jt. Surg., vol. 61, no. 5, pp. 756–759, 1979, doi: [91] I. W. Hunter and S. Lafontaine, “Comparison of muscle with artificial
10.2106/00004623-197961050-00017. actuators,” in Technical Digest- IEEE Solid-State Sensor and Actuator
[68] M. Sardelli, R. Z. Tashjian, and B. A. MacWilliams, “Functional elbow Workshop, 1992, pp. 178–185, doi: 10.1109/solsen.1992.228297.
range of motion for contemporary tasks,” J. Bone Jt. Surg., vol. 93, no. 5, [92] R. Damerla and S. Awtar, “Constraint-Based Analysis of Parallel
pp. 471–477, 2011, doi: 10.2106/JBJS.I.01633. Kinematic Articulated Wrist Mechanisms,” J. Mech. Robot., vol. 13, no.
[69] B. F. Morrey, L. J. Askew, K. N. An, and E. Y. Chao, “A biomechanical 3, Jun. 2021, doi: 10.1115/1.4049947.
study of normal functional elbow motion,” J. Bone Jt. Surg., vol. 63, no. [93] Y. D. Patel and P. M. George, “Parallel Manipulators Applications-A
6, pp. 872–877, 1981, doi: 10.2106/00004623-198163060-00002. Survey,” Mod. Mech. Eng., vol. 2, pp. 57–64, 2012, doi:
[70] S. Henmi, K. Yonenobu, T. Masatomi, and K. Oda, “A biomechanical 10.4236/mme.2012.23008.
study of activities of daily living using neck and upper limbs with an [94] G. Pritschow, “Parallel Kinematic Machines (PKM) – Limitations and
optical three-dimensional motion analysis system,” Mod. Rheumatol., vol. New Solutions,” Ann. of the CIRP, vol. 49, no. 1, pp. 275–280, Jan. 2000,
16, pp. 289–293, 2006, doi: 10.1007/s10165-006-0499-x. doi: 10.1016/S0007-8506(07)62945-X.
[71] J. Aizawa et al., “Three-dimensional motion of the upper extremity joints [95] J. L. Pons et al., “The MANUS-HAND * Dextrous Robotics Upper Limb
during various activities of daily living,” J. Biomech., vol. 43, no. 15, pp. Prosthesis: Mechanical and Manipulation Aspects,” Auton. Robots, vol.
2915–2922, 2010, doi: 10.1016/j.jbiomech.2010.07.006. 16, pp. 143–163, 2004.
[72] D. L. Nelson, M. A. Mitchell, P. G. Groszewski, S. L. Pennick, and P. R. [96] M. L. Davidson, “Development of a Novel Prosthetic Wrist Device
Manske, “Wrist Range of Motion in Activities of Daily Living,” Adv. Incorporating the Dart Thrower’s Motion,” Ph.D. dissertation, Bioeng.,
Biomech. Hand Wrist, pp. 329–334, 1994, doi: 10.1007/978-1-4757- University of Colorado at Denver, Denver, CO, USA, 2017.
9107-5_29. [97] G. Lee, G. Y. Hong, and Y. Choi, “Tendon-driven compliant prosthetic
[73] D. H. Gates, L. S. Walters, J. Cowley, J. M. Wilken, and L. Resnik, wrist consisting of three rows based on the concept of tensegrity
“Range of motion requirements for upper-limb activities of daily living,” structure,” IEEE Robot. Autom. Lett., vol. 6, no. 2, pp. 3956–3963, 2021,
Am. J. Occup. Ther., vol. 70, no. 1, pp. 1–10, 2016, doi: doi: 10.1109/LRA.2021.3067237.
10.5014/ajot.2016.015487. [98] P. J. Kyberd et al., “Two-degree-of-freedom powered prosthetic wrist,” J.
[74] J. Ryu, W. P. Cooney, L. J. Askew, K. N. An, and E. Y. S. Chao, Rehabil. Res. Dev., vol. 48, no. 6, pp. 609–618, 2011, doi:
“Functional ranges of motion of the wrist joint,” J. Hand Surg. Am., vol. 10.1682/JRRD.2010.07.0137.
16, no. 3, pp. 409–419, 1991, doi: 10.1016/0363-5023(91)90006-W. [99] N. M. Bajaj and A. M. Dollar, “Design and Preliminary Evaluation of a
[75] M. Doğan et al., “Functional range of motion in the upper extremity and 3-DOF Powered Prosthetic Wrist Device,” in Proceedings of the IEEE
trunk joints: Nine functional everyday tasks with inertial sensors,” Gait RAS and EMBS International Conference on Biomedical Robotics and
Posture, vol. 70, pp. 141–147, 2019, doi: 10.1016/j.gaitpost.2019.02.024. Biomechatronics, 2018, pp. 119–125, doi:
[76] W. Dempster and G. Gaughran, “Properties of Body Sements Based on 10.1109/BIOROB.2018.8487192.
Size and Weight,” Am. J. Anat., vol. 120, pp. 33–54, 1967. [100] R. Mahmoud, A. Ueno, and S. Tatsumi, “An assistive tele-operated
[77] S. Plagenhoef, F. Gaynor Evans, and T. Abdelnour, “Anatomical Data for anthropomorphic robot hand: Osaka city university hand II,” in 2011 6th
Analyzing Human Motion,” Res. Quarterly Exerc. Sport, vol. 54, no. 2, ACM/IEEE International Conference on Human-Robot Interaction
pp. 169–178, 1983, doi: 10.1080/02701367.1983.10605290. (HRI), 2011, pp. 85–92, doi: 10.1145/1957656.1957677.
[78] C. E. Clauser, J. T. McConville, and J. W. Young, “Weight, Volume, and [101] N. A. Abd Razak, N. A. Abu Osman, H. Gholizadeh, and S. Ali,
Center of Mass of Segments of the Human Body,” Natl. Tech. Inf. Serv., “Development and performance of a new prosthesis system using
pp. 1–101, 1969, doi: AMRL-TR- 69-70 (AD 710 622). ultrasonic sensor for wrist movements: A preliminary study,” Biomed.
[79] P. de Leva, “Adjustments to Zatsiorsky-Selyanov’s Segment Inertia Eng. Online, vol. 13, pp. 1–14, 2014, doi: 10.1186/1475-925X-13-49.
Parameters,” J. Biomech., vol. 29, no. 9, pp. 1223–1230, 1996. [102] J. Lipsey and J. Sensinger, “Modular and lightweight myoelectric
[80] J. Shealy and W. Latko, “Effects of Mass on Wrist Velocities and prosthesis components and related methods,” U.S. Patent 2019/0380846
Accelerations,” in Proceedings of the Human Factors Society 36th A1, 2019.
Annual Meeting, 1992, pp. 761–764. [103] T. Lenzi, J. Lipsey, and J. W. Sensinger, “The RIC Arm - A Small
[81] H. Ida, K. Fukuhara, S. Kusubori, and M. Ishii, “A study of kinematic Anthropomorphic Transhumeral Prosthesis,” IEEE/ASME Trans.
cues and anticipatory performance in tennis using computational Mechatronics, vol. 21, no. 6, pp. 2660–2671, 2016, doi:
manipulation and computer graphics,” Behav. Res. Methods, vol. 43, pp. 10.1109/TMECH.2016.2596104.
781–790, 2011, doi: 10.3758/s13428-011-0084-x. [104] N. R. Olsen et al., “An Adaptable Prosthetic Wrist Reduces Subjective
[82] S. K. Charles and N. Hogan, “Dynamics of wrist rotations,” J. Biomech., Workload,” bioRxiv, Oct. 2019, doi: 10.1101/808634.
vol. 44, no. 4, pp. 614–621, 2011, doi: 10.1016/j.jbiomech.2010.11.016. [105] C. Roose, “Two-Degree-of-Freedom Pneumatically Powered Wrist
[83] G. Å. Hansson et al., “Physical workload in various types of work: Part I. Prosthesis,” Delft University of Technology, Delft, Netherlands, 2014.
Wrist and forearm,” Int. J. Ind. Ergon., vol. 39, no. 1, pp. 221–233, 2009, [106] H. Takeda, N. Tsujiuchi, T. Koizumi, H. Kan, M. Hirano, and Y.
doi: 10.1016/j.ergon.2008.04.003. Nakamura, “Development of prosthetic arm with pneumatic prosthetic
[84] L. W. O’Sullivan and T. J. Gallwey, “Forearm torque strengths and hand and tendon-driven wrist,” in Proceedings of the 31st Annual
discomfort profiles in pronation and supination,” Ergonomics, vol. 48, no. International Conference of the IEEE Engineering in Medicine and
6, pp. 703–721, 2005, doi: 10.1080/00140130500070954. Biology Society: Engineering the Future of Biomedicine, EMBC 2009,
[85] J. Matsuoka, R. A. Berger, L. J. Berglund, and K. An, “An Analysis of of 2009, pp. 5048–5051, doi: 10.1109/IEMBS.2009.5333668.
the Forearm Under Resisted Forearm Rotation in Normal Subjects,” J. [107] M. Hioki et al., “Design and control of electromyogram prosthetic hand
Hand Surg. Am., vol. 31A, no. 5, pp. 801–805, 2006. with high grasping force,” in 2011 IEEE International Conference on
[86] P. Axelsson, P. Fredrikson, A. Nilsson, J. K. Andersson, and J. Kärrholm, Robotics and Biomimetics, ROBIO 2011, 2011, pp. 1128–1133, doi:
“Forearm Torque and Lifting Strength: Normative Data,” J. Hand Surg. 10.1109/ROBIO.2011.6181439.
Am., vol. 43, no. 7, pp. 677.e1-677.e17, 2018, doi: [108] H. Kawasaki, T. Mouri, T. Hara, and H. Shimomura, “Humanoid
10.1016/j.jhsa.2017.12.022. Electric Hand,” U.S. Patent 8,747,486 B2, 2014.
[87] Y. Yoshii, H. Yuine, O. Kazuki, W. lin Tung, and T. Ishii, “Measurement [109] K. B. Fite, T. J. Withrow, X. Shen, K. W. Wait, J. E. Mitchell, and M.
IEEE Transactions on Medical Robotics and Bionics 17

Goldfarb, “A gas-actuated anthropomorphic prosthesis for transhumeral “Architecture for the semi-automatic fabrication and assembly of thin-
amputees,” IEEE Trans. Robot., vol. 24, no. 1, pp. 159–169, Feb. 2008, film based dielectric elastomer actuators,” Proc. SPIE 6927,
doi: 10.1109/TRO.2007.914845. Electroactive Polymer Actuators and Devices (EAPAD) 2008, vol. 6927,
[110] S. K. Kundu and K. Kiguchi, “Development of a 5 DOF prosthetic arm pp. 739–748, Apr. 2008, doi: 10.1117/12.784981.
for above elbow amputees,” in Proceedings of 2008 IEEE International [131] P. Rothemund et al., “HASEL Artificial Muscles for a New Generation
Conference on Mechatronics and Automation, ICMA 2008, 2008, pp. of Lifelike Robots—Recent Progress and Future Opportunities,” Adv.
207–212, doi: 10.1109/ICMA.2008.4798753. Mater., vol. 33, no. 19, p. 2003375, May 2021, doi:
[111] F. Zajac, “Muscle and tendon: properties, models, scaling, and 10.1002/ADMA.202003375.
application to biomechanics and motor control.,” Crit. Rev. Biomed. [132] Z. Yoder et al., “Design of a High-Speed Prosthetic Finger Driven by
Eng., vol. 17, no. 4, pp. 359–410, 1989. Peano-HASEL Actuators,” Front. Robot. AI, vol. 7, p. 181, Nov. 2020,
[112] “Pneumatic cylinders | McMaster-Carr.” [Online]. Available: doi: 10.3389/FROBT.2020.586216/BIBTEX.
https://www.mcmaster.com/pneumatic-cylinders/. [133] N. Kellaris, V. G. Venkata, P. Rothemund, and C. Keplinger, “An
[113] W. Liang, H. Liu, K. Wang, Z. Qian, L. Ren, and L. Ren, “Comparative analytical model for the design of Peano-HASEL actuators with
study of robotic artificial actuators and biological muscle,” Adv. Mech. drastically improved performance,” Extrem. Mech. Lett., vol. 29, p.
Eng., vol. 12, no. 6, p. 168781402093340, Jun. 2020, doi: 100449, May 2019, doi: 10.1016/J.EML.2019.100449.
10.1177/1687814020933409. [134] R. Hinchet, H. Shea, R. Hinchet, and H. Shea, “High Force Density
[114] I. W. Hunter, J. M. Hollerbach, and J. Ballantyne, “A comparative Textile Electrostatic Clutch,” Adv. Mater. Technol., vol. 5, no. 4, p.
analysis of actuator technologies for robotics,” Robot. Rev. 2, pp. 299– 1900895, Apr. 2020, doi: 10.1002/ADMT.201900895.
342, 1991. [135] F. Claeyssen and N. Lhermet, “Actuators Based on Giant
[115] A. Nespoli, S. Besseghini, S. Pittaccio, E. Villa, and S. Viscuso, “The Magnetostrictive Materials,” in ACTUATOR 2002, 8th International
high potential of shape memory alloys in developing miniature Conference on New Actuators, 2002, pp. 148–153.
mechanical devices: A review on shape memory alloy mini-actuators,” [136] G. Dai et al., “A review of magnetostrictive iron–gallium alloys,” Smart
Sensors Actuators A Phys., vol. 158, no. 1, pp. 149–160, Mar. 2010, doi: Mater. Struct., vol. 20, no. 4, p. 043001, Mar. 2011, doi: 10.1088/0964-
10.1016/J.SNA.2009.12.020. 1726/20/4/043001.
[116] H. Stroud and D. Hartl, “Shape memory alloy torsional actuators: a [137] N. Gabdullin and S. H. Khan, “Review of properties of magnetic shape
review of applications, experimental investigations, modeling, and memory (MSM) alloys and MSM actuator designs,” J. Phys. Conf. Ser.,
design,” Smart Mater. Struct., vol. 29, no. 11, p. 113001, Oct. 2020, doi: vol. 588, no. 1, p. 012052, Feb. 2015, doi: 10.1088/1742-
10.1088/1361-665X/ABBB12. 6596/588/1/012052.
[117] J. Mohd Jani, M. Leary, A. Subic, and M. A. Gibson, “A review of shape [138] J. Tellinen, I. Suorsa, A. Jääskeläinen, I. Aaltio, and K. Ullakko, “Basic
memory alloy research, applications and opportunities,” Mater. Des., Properties of Magnetic Shape Memory Actuators,” in ACTUATOR
vol. 56, pp. 1078–1113, Apr. 2014, doi: 2002, 8th International Conference on New Actuators, 2002.
10.1016/J.MATDES.2013.11.084. [139] “Maxon Selection Guide: 2019/2020,” Maxon Group, Sachseln,
[118] J. Mohd Jani, M. Leary, and A. Subic, “Designing shape memory alloy Switzerland. [Online]. Available:
linear actuators: A review:,” https://www.maxongroup.co.uk/maxon/view/news/The-new-maxon-
http://dx.doi.org/10.1177/1045389X16679296, vol. 28, no. 13, pp. catalogue-for-20192020
1699–1718, Dec. 2016, doi: 10.1177/1045389X16679296. [140] “Faulhaber Drive Systems: 2018,” Faulhaber Micromo, Clearwater, FL,
[119] K. Andrianesis and A. Tzes, “Development and Control of a USA. [Online]. Available: https://pdf.directindustry.com/pdf/faulhaber-
Multifunctional Prosthetic Hand with Shape Memory Alloy Actuators,” drive-systems-7023.html
J. Intell. Robot. Syst., vol. 78, pp. 257–289, May 2015, doi: [141] “Adamant Namiki Customized Actuator Catalog,” Adamant Namiki
10.1007/s10846-014-0061-6. Precision Jewel Co., Tokyo, Japan. [Online]. Available: https://www.ad-
[120] “Cedrat Technologies Product Catalog,” Cedrat Technologies, Meylan, na.com/en/product/dccorelessmotor/brushlessmotor.html
France. [Online]. Available: https://www.cedrat- [142] “Brushless DC Motors From ElectroCraft,” ElectroCraft, Inc., Stratham,
technologies.com/en/download-1.html NH, USA. [Online]. Available:
[121] “PiezoMove High-Stiffness Linear Piezo Actuator,” PI USA, Auburn, https://www.electrocraft.com/products/bldc/.
MA, USA. [Online]. Available: https://www.pi- [143] “Moog Brushless Motors Overview,” Moog Inc., Elma, NY, USA.
usa.us/en/products/piezo-actuators-stacks-benders-tubes/p-602- [Online]. Available:
piezomove-high-stiffness-linear-piezo-actuator-202700/#specification https://www.moog.com/content/sites/global/en/products/motors-
[122] “DSM Piezo Actuators,” Dynamic Structures & Materials, LLC, servomotors/brushless-motors/.
Franklin, TN, USA. [Online]. Available: https://www.dynamic- [144] “MinebeaMitsumi Brushless Motors,” MinebeaMitsumi Inc., Tokyo,
structures.com/piezo-actuators Japan. [Online]. Available:
[123] A. Shafik and R. Ben Mrad, “Piezoelectric Motor Technology: A https://www.minebeamitsumi.eu/en/brushless-motor/.
Review,” Nanopositioning Technol. Fundam. Appl., pp. 33–59, Jan. [145] “Harmonic Drive General Catalog”, Harmonic Drive LLC, Beverly,
2016, doi: 10.1007/978-3-319-23853-1_2. MA, USA. [Online]. Available:
[124] G. Rui Zhang, P. Lochmatter, A. Kunz, G. M. Kovacs, R. Zhang, and G. https://www.harmonicdrive.net/downloads/catalogs
Kovacs, “Spring roll dielectric elastomer actuators for a portable force [146] “Cone Drive Harmonic Solutions Catalog,” Cone Drive, Traverse City,
feedback glove,” https://doi.org/10.1117/12.658524, vol. 6168, no. 22, MI, USA. [Online]. Available: https://conedrive.com/resources/
pp. 505–516, Mar. 2006, doi: 10.1117/12.658524. [147] “GAM GSL Strain Wave Gearbox Catalog,” GAM, Prospect, IL, USA.
[125] R. W. Jones et al., “An arm wrestling robot driven by dielectric [Online]. Available: https://www.gamweb.com/gsl-robotic-strain-
elastomer actuators,” Smart Mater. Struct., vol. 16, no. 2, p. S306, Mar. wave-gearbox.html
2007, doi: 10.1088/0964-1726/16/2/S16. [148] “DieQua Precision Cycloidal Reducers Catalog,” DieQua Corporation,
[126] T. Lu, C. Ma, and T. Wang, “Mechanics of dielectric elastomer Bloomingdale, IL, USA. [Online]. Available:
structures: A review,” Extrem. Mech. Lett., vol. 38, p. 100752, Jul. 2020, https://diequa.com/catalog-downloads/
doi: 10.1016/J.EML.2020.100752. [149] “Nabtesco Precision Reduction Gear RV Catalog,” Nabtesco Motion
[127] T. Lu, C. Chiang Foo, J. Huang, J. Zhu, and Z. Suo, “Highly deformable Control, Inc., Farmington Hills, MI, USA. [Online]. Available:
actuators made of dielectric elastomers clamped by rigid rings,” J. Appl. https://www.nabtescomotioncontrol.com/downloads/
Phys, vol. 115, no. 18, p. 184105, May 2014, doi: 10.1063/1.4876722. [150] “KHK Worm Gear Pair Catalog,” Kohara Gear Industry Co., Ltd.,
[128] G. Kovacs, L. Düring, S. Michel, and G. Terrasi, “Stacked dielectric Kawaguchi, Saitama, Japan. [Online]. Available:
elastomer actuator for tensile force transmission,” Sensors Actuators A: https://khkgears.net/new/worm_gear.html
Physical, vol. 155, no. 2, pp. 299–307, Oct. 2009, doi: [151] “SDP/SI Worms & Worm Gears Catalog,” Designatronics Inc.,
10.1016/J.SNA.2009.08.027. Hicksville, NY, USA. [Online]. Available: https://www.sdp-
[129] D. Rossi, D. F. Carpi, D. De Rossi, and F. Carpi, “Contractile folded si.com/products/Gears/Index.php
dielectric elastomer actuators,” Proc. SPIE 6524, Electroactive Polymer [152] “Rush Gears Worm Gear Ratio & Torque Info,” Rush Gears Inc., Fort
Actuators and Devices (EAPAD) 2007, vol. 6524, no. 4, pp. 127–139, Washington, PA, USA. [Online]. Available:
Apr. 2007, doi: 10.1117/12.715594. https://www.rushgears.com/tech-tools/worm-gear-ratios-and-torque.
[130] M. Randazzo, R. Buzio, G. Metta, G. Sandini, and U. Valbusa, [153] “KSS Standard Products of Ball Screws,” KSS Co., Ltd., Ohta-ku,
IEEE Transactions on Medical Robotics and Bionics 18

Tokyo, Japan. [Online]. Available:

https://www.kssballscrew.com/us/download/download.html
[154] “NSK Ball Screws,” NSK Ltd., Shinagawa-Ku, Tokyo, Japan. [Online].
Available: https://www.nsk.com/products/precisionmachine/ballscrew/
[155] “Thomson Ball Screws and Ball Splines,” Thomson Industries, Inc.,
Radford, VA, USA. [Online]. Available:
https://www.thomsonlinear.com/en/products/ball-screws#literature
[156] “TBI Motion Ball Screw Catalog,” TBI Motion Technology Co., Ltd.,
New Taipei City, Taiwan. [Online]. Available:
https://www.tbimotion.com.tw/en/product/Ball-Screw-SFK/ball-
screw_SFK.html
[157] “Misumi Motion Systems Product Catalog,” MISUMI Corporation,
Schaumberg, IL, USA. [Online]. Available: https://us.misumi-
ec.com/contents/company/about/
[158] “SDP/SI Timing belts, Pulleys, Chains and Sprockets Catalog,”
Designatronics Inc., Hicksville, NY, USA. [Online]. Available:
https://www.sdp-si.com/products/Gears/Index.php
[159] “KSS Precision Lead Screws,” KSS Co., Ltd., Ohta-ku, Tokyo, Japan.
[Online]. Available:
https://www.kssballscrew.com/us/download/download.html
[160] “Roller Chain,” USA Roller Chain & Sprockets, Winter Garden, FL,
USA. [Online]. Available: https://www.usarollerchain.com/Roller-
Chain-s/2115.htm.
[161] S. Bai, X. Li, and J. Angeles, “A review of spherical motion generation
using either spherical parallel manipulators or spherical motors,” Mech.
Mach. Theory, vol. 140, pp. 377–388, Oct. 2019, doi:
10.1016/J.MECHMACHTHEORY.2019.06.012.
[162] “Hand Chassis w/ Quick Disconnect Wrist.” [Online]. Available:
https://www.spshangerstore.com/hand-chassis-w-quick-disconnect-
wrist.html.
[163] L. F. Bieg and G. L. Benavides, “Large displacement spherical joint,”
U.S. Patent 6,409,413 B1, 2002.
[164] N. M. Bajaj and A. M. Dollar, “Design of a Large Workspace Passive
Spherical Joint via Contact Edge Design,” Proc. ASME Des. Eng. Tech.
Conf., vol. 10, Nov. 2020, doi: 10.1115/DETC2020-22753.
[165] A. F. Azocar, L. M. Mooney, L. J. Hargrove, and E. J. Rouse, “Design
and Characterization of an Open-Source Robotic Leg Prosthesis,” Proc.
IEEE RAS EMBS Int. Conf. Biomed. Robot. Biomechatronics, vol. 2018-
August, pp. 111–118, Oct. 2018, doi: 10.1109/BIOROB.2018.8488057.