Abstract

Existing devices to assist upper extremity (UE) movement in infants with or at risk for motor impairments remain limited and are mainly passive devices. The aim of this project was to develop and assess the validity and reliability of the first-actuated wearable device for this population. A wearable device consisting of four pneumatic actuators (two per arm) was developed and tested on a custom-built physical model with articulated joints (four degrees-of-freedom (DOFs) per arm) based on an average 12-month-old infant's upper body. The device actively controls 2DOFs per arm (one at the elbow and one at the shoulder) and does not prohibit motion about the remaining non-actuated DOFs. Three distinct device actuator synergies, that resemble muscle recruitment strategies, were evaluated in a vertical reaching task using one arm and both arms. The device was assessed for its performance, wearability, and safety. Performance was assessed via the average duration, smoothness, and repeatability of reaching movements, and maximum range of motion per actuated joint. Wearability was assessed via kinematic compatibility to infant reaching trajectories. Safety was assessed via actuator durability. Results demonstrate the efficacy of the device and reveal key insights for further improvements.

1 Introduction

There is an urgent need to develop technology for diagnosis, intervention, and assessment of children with brain injury and motor disability [1]. Brain injury and developmental disabilities are highly prevalent [2,3] and often lead to impaired upper extremity (UE) motor function. This, in turn, leads to limited motor experiences early in life that affect other developing systems such as perception and cognition [4,5]. Given that UE motor skills largely develop over the course of the first 6 months of life (e.g., reaching, object exploration, etc.) [6] and the majority of brain changes take place in the first 2 years of life [7], an early start to intervention (i.e., while at infancy) is critical. For this population, gaining early motor experiences may require the use of assistive technology.

Existing types of assistive technology to support UE movement in infants remain limited and mainly comprise rigid exoskeletons and soft wearable devices. The Pediatric Wilmington Robotic Exoskeleton (P-WREX) was the first UE-assistive device developed for and tested with infants with motor impairments. The P-WREX is a portable three-dimensional (3D)-printed, two-link structure mounted on a body brace that is passively actuated by elastic bands placed on the forearm and the upper arm links. A case study documented the feasibility and preliminary effectiveness of the P-WREX after long-term, in-home use, and training in an infant with movement impairments [8]. This exoskeleton design has certain advantages, such as the user-friendly force manipulation through the placement of rubber bands and the implementation of small rigid components to provide adequate mechanical support [9]. This design, however, also comes with several functional and esthetic limitations (due to bulkiness and rigid components) that may affect physical and social interactions within the environment. Soft wearable devices are designed to address these limitations. A wearable design is different in that it uses the human body as a support environment for the device, and thus, it respects the body's changing and diverse form [10]. The Playskin Lift was the first soft wearable UE device developed for children younger than 3 year of age. It is a garment made of four-way stretch blended fabric that contains under-arm vinyl tunnels in either side of the trunk [11]. Playskin Lift is passively actuated by mechanical inserts that are placed into the tunnels. Longitudinal training with this device leads to an improvement in UE motor abilities of infants with and at risk for UE motor impairments [12,13].

Both types of technology, while successful in assisting and training UE movement in infants, share a common limitation; they passively and continuously offer the same level of assistance. Thus, an adult has to manually change parts of the device (e.g., swap rubber bands or mechanical inserts of different stiffness) to alter the equilibrium point of the combined user–device system. Consequently, the level of support cannot be adjusted in real-time to meet the evolving biomechanical demands for support within a task (i.e., different phases of reaching) or across tasks.

Recently, a user-controlled prototype using artificial muscles to actively assist older children with lifting their arms against gravity was developed [14]. However, the device requires grasping control from the child to inflate and deflate an under-arm artificial muscle, and thus, it cannot be controlled directly by an infant. Therefore, designing adaptive wearable UE devices that can accurately detect, measure, inform, and adapt to slow and subtle changes in motor learning and growth in infancy still remains a challenge [1].

Developing technology for infants extends beyond the redesign and application of miniaturized versions of successful assistive devices for older children and adults. The performance of such device should align with the movement capabilities of the developing infants and training goals for this population. For example, the kinematic parameters of infants' reaching motion are different from those of adults [15]. In addition, temporal multijoint coordination during the first three years of life is highly variable. Movement variability is essential for the identification of movement strategies that satisfy the objective of reaching for a presented object [1517]. Thus, it is important for a pediatric device to allow young populations experience movement variability (and not just the most optimal movement path) by providing various kinematic strategies similar to those seen in infant UE motor learning.

Achieving kinematic compatibility between the motion induced by a device and that of human joints is only one aspect of technology design for wearability. Other aspects involve identification of the appropriate placement, sizing, and form of attachment on the body [10,18]. Additional common design and manufacturing considerations for UE wearable devices are safety and performance [19]. Design for performance may refer to the ability of the device to execute the tasks in a timely and accurate manner, among others [19]. Design for safety may refer to taking precautions to prevent extra musculoskeletal injuries due to excessive movement (e.g., motion beyond joint range, excessive joint velocities, etc.) [20].

The goal of this work was to develop the first-actuated UE wearable device for infants and assess its performance, wearability, and safety in a reaching task. Reaching was selected as it is a precursor to environment exploration and learning in early human development and a major focus in pediatric UE intervention [21,22]. For this initial assessment of the first prototype, all experiments were conducted with a custom-made physical infant model. Data from this bench testing would determine the feasibility of the next step, which is testing with infants with UE motor impairments.

2 Materials and Methods

2.1 Prototype Development

2.1.1 Design Approach.

The prototype system developed in this work includes two main components: the actual soft robotic wearable suit (see Fig. 1), and an off-suit pneumatic control board which is placed away from the body for safety. The suit is a stretchable fabric garment (off-the-shelf elastic apparel) with integrated pneumatic actuators and air tunnels. There are four pneumatic actuators in total (two per arm). Each actuator supports 1DOF motion about a specific joint. One actuator is responsible for shoulder abduction to help lift the arms against gravity. The other is responsible for elbow flexion. It is important to note that while the wearable device actuates 2DOFs in each arm, it does not prohibit motion about the remaining nonactuated DOFs of the arms. For example, the infant may concurrently perform shoulder abduction (which is actuated by the device) and horizontal flexion (self-actuated by muscles), if some residual muscle activity is present. This human-machine muscle cooperation may be vital for motor training and avoidance of sole reliance on the device.

2.1.2 Actuation.

A range of soft actuators were manufactured (via 3D printing and casting) and tested to identify the actuator size and shape characteristics appropriate for a device intended for use by infants. Traditional electromechanical actuators like motors, servos, and pistons were avoided to: (i) reduce the risk of bodily injury because of their rigidity [23], and (ii) keep the additional weight low (<10% of arm weight) and ensure the quantity and quality of infants' movements are not altered [24]. In addition, initial testing revealed that casted actuators are more suitable for our application than the 3D-printed ones. Specifically, pneumatic actuators inspired by existing designs, such as the McKibben artificial muscles [25] and the PneuNets bending and fiber-reinforced actuators [2630], combine the benefits of small size and low weight with generation of the required for our application forces. Anthropometric data suggest that the upper arm length varies on average between 13.8 and 18.7 cm for male and 13.5 and 18.4 cm for female infants in the age range of 6 months to 2 years old [31]. For the same age range, the average total arm weight is between 0.3 and 0.6 kg. Thus, the group of soft actuators at each arm needs to be able to generate a combined force of at least 3–6 N, for the device to offer sufficient antigravity support. Casted actuators are capable of generating that amount of support.

Two distinct types of casted actuators were placed at the shoulder and elbow joints. A custom-made elongated-bellow-like soft actuator was employed for the shoulder abduction/adduction (see left panel of Fig. 2). For the elbow flexion/extension, a PneuNets bending actuator was retrofitted [26] (see right panel of Fig. 2). These actuators use pneumatic pressurization and depressurization (or sometimes vacuum) applied to deformable chambers or bellows to create linear or rotational motion. Casting molds were 3D-printed via a Mark Two 3D printer using carbon-fiber-reinforced Onyx filament, while all soft actuators were made from Smooth-On Dragon Skin 30 silicone.

Identifying appropriate anchor points for the actuators is critical. The shoulder actuator was placed under the armpit with one end anchored laterally at the midtorso and the other end anchored at the side midpoint of the upper arm. The elbow actuator was anchored on the rear upper arm and on the joint elbow joint. The other end was allowed to slide upon the forearm; this is needed to give the freedom to the actuator to curl and push the forearm as it is being pressurized. It is important to keep in mind that the anchor points are directly related to the size and motion and force generation profiles of the specific actuator used. At this stage of development, this initial “calibration” phase is required. In this first prototype, we used velcro straps to anchor the actuators.

2.1.3 Pneumatic Control.

A pneumatic control board was built that contains all the key components for controlling the wearable device, such as miniaturized diaphragm pumps, solenoid valves, and a microcontroller (Arduino Mega). In detail, the board contains two pumps; each pump pressurizes two air channels that lead to the pneumatic actuators of a single arm. Eight solenoid valves direct air flow; each valve is regulated through pulse-width modulation (PWM) from the microcontroller. Three-way toggle switches are used to manually choose between pressurization, depressurization, and ventilation modes. A set of MOSFETs are used to read the desired PWM and regulate voltage at the solenoid valves. Furthermore, one additional MOSFET controls the ON/OFF states for the pneumatic pumps. Finally, four potentiometers are used to change the duty cycle of PWM, which in turn controls the opening/closure of valves. Figure 3 depicts the interconnection of the various components of our pneumatic control board. The physical prototype is featured in Fig. 4.

2.2 Experiments.

Device performance, wearability, and safety were assessed through a series of experiments that involved the placement of our wearable device on a physical wooden model. The model was scaled to approximate a 50th percentile 12-month-old infant's upper body [31]. The wooden model (featured in Fig. 4) consists of a rigid torso, two rigid upper arm parts, and two rigid forearm parts. The total weight of each arm is 0.45 kg. The upper arms are connected to the torso through 3DOFs ball-and-socket shoulder joints. The forearms are connected to the upper arms through 1DOF hinge-elbow joints. Thus, each arm-torso subsystem has 4DOFs. Two of these (one in shoulder and the other in the elbow) are controlled via the pneumatic actuators of the device.

During experiments, the wearable device was actuated to lift both of the model's arms against gravity and lead their tips (i.e., end-effectors) toward a virtual object placed up front and at the model's chest height. Actuation was implemented using three different “muscle” recruitment strategies. The “muscle” recruitment strategies were selected to elicit different patterns of temporal coordination across arm segments, similar to those seen during the development of infant reaching. Previous work on reaching a single target within an infant's workspace has identified the following patterns: Newly reaching infants' arm segments move independently to reach, with elbow extension preceding or following shoulder abduction. However, over time, and by 24 months, a more stable interjoint synergy emerges where shoulder abduction initiates the reach which is then followed by elbow extension [15].

Accordingly, we considered the aforementioned sequential actuation strategies along with one simultaneous actuation strategy to perform reaching movements. More specifically:

  • The simultaneous (Sim) strategy involved simultaneous pressurization of both actuators of each arm, producing simultaneous shoulder abduction and elbow flexion.

  • The shoulder-elbow (ShEl) sequential strategy included pressurization of the shoulder actuator first and of the elbow second.

  • The elbow-shoulder (ElSh) sequential strategy included pressurization of the elbow actuator first and of the shoulder actuator second.

We tested each strategy under three different pressurization conditions (controlling the valves at 25%, 50%, and 75% duty cycle) with one arm (for a subtotal of 45 trials) and with both arms (for a subtotal of 90 trials). The arms can return to the home position either through passive or controlled depressurization. The passive assistance that our device is able to provide this way may enable the infant to self-initiate movement with some remaining passive assistance, which is an important feature to include for their motor learning.

Three-dimensional position and orientation data of the model's end-effector were collected via motion capture (12-camera Vicon Vantage V5 and Vero2.2 at 100 Hz). End-effector motion trajectories were obtained using each strategy and analyzed. Performance was assessed by looking at the (i) average duration of reaching movements (temporal difference from initiation of movement until convergence to a steady-state maximum distal point), (ii) trajectory smoothness (integral of the square of the second-order derivative), (iii) repeatability of reaching movements between different repetitions (through intra-class correlation coefficient and coefficient of variation), and (iv) maximum range of motion per joint (measured about the actuated axis of rotation). Wearability was assessed by looking at the kinematic compatibility between the trajectories obtained herein from the actuated device and infant reaching trajectories identified in previous work through visual inspection and comparison. Safety was assessed by looking at the actuator durability (the number of reaching movements each actuator performed before failure) as well as the type of failure.

3 Results

We demonstrate 3D reaching trajectories for both arms (Fig. 5) and for a single arm (Fig. 6). End-effector 3D position data of the single-arm case for every “muscle” recruitment strategy are depicted in Figs. 79.

3.1 Performance.

Average duration of reaching movements varied between 2.36 s and 4.93 s depending on the employed strategy (see Table 1). The Sim strategy produced the fastest movements as shown by the average duration across all three duty cycles (2.39 s). It is followed by the ShEl strategy (3.85 s), and then by the ElSh strategy (4.73 s). The effect of the duty cycle on duration of reaching movements seems to be minimal, with larger variations at the ElSh strategy.

Overall, trajectories were observed to be quite smooth, as it can be visually confirmed in Figs. 79. The employed smoothness coefficient reveals the relative smoothness of each strategy, as a function of the duty cycle (Table 2). In the table, we show the smoothness values about the X-, Y-, and Z-components of the single-arm reaching trajectories. Smaller numbers indicate more smooth trajectories. The ElSh strategy is found to lead to smoother trajectories. The finding may be explained by the observation that this strategy was also the slowest to converge to its maximum distal point (Table 1). The ShEl and then Sim strategies are found comparatively less smooth. Further, results suggest that the duty cycle may impact the smoothness of the output trajectories. Smaller duty cycles seem to lead to comparatively smoother trajectories, for every strategy employed. It can be also observed that the Z-component of all trajectories have higher values. This may be explained by the fact that the majority of work done by the device is allocated toward raising the arms, i.e., along the Z-direction; less work is needed to move the end-effector along the X- and Y-directions.

Repeatability of motion measurements with the device was high (ICC over 0.9 on average) using any of the three strategies (Table 3). In the table, we show the ICC values about the X-, Y-, and Z-components of the single-arm reaching trajectories. It can be observed that repeatability about the Z-component is the highest, followed by repeatability about the Y- and X-components. This finding is line with component-wise trajectory smoothness findings (Table 2), suggesting that the device, at its current actuator configuration, can better control the arm about the Z-direction. The aforementioned observations are further supported by the coefficient of variation among trajectory components (Table 4).

In all cases, the maximum angle achieved by actuation in both shoulder and elbow joints was approximately 90 deg.

3.2 Wearability.

It can be readily verified (Fig. 5) that the three “muscle” recruitment strategies lead to distinct trajectories, with the ElSh being the most distinctive one, while the other two strategies resemble more each other. For clarity purposes, single-arm 3D curves for each case are shown in Fig. 6.

The trajectories obtained in our experiments resemble reaching trajectories typically seen early on, such as in 5 and 9 month-old infants (cf. [15, Fig. 2]). No straight paths were observed (Fig. 6). This is because the current wearable prototype can actuate 2DOFs per arm (although the device itself does not prohibit motion along the remaining nonactuated degrees-of-freedom). Future work focusing on integrating actuation for more DOFs may help generate more straight trajectories typically seen in 15 and 24 month-old infants [15]. Eventually, we hope to gain a wide range of trajectories using different muscle recruitment strategies that correspond to the wide range of trajectories typically seen in infants as they learn how to reach over time.

3.3 Safety.

We report the number of repetitions the device's actuators were able to achieve before failing. These data were collected during both arms reaching experiments. We used a total of two elbow and five shoulder actuators. Out of the two types of actuators we used, the retrofitted PneuNet (elbow) actuators appear to have higher durability that the elongated-bellow (shoulder) actuators (see Table 5).

Testing revealed the most critical points of failure for each actuator. Retrofitted PneuNet actuators were primarily failing at the side of a groove. One way to remedy this is by a slight increase of the depth of each groove's sides. Bellow actuators were consistently failing at their middle where the upper and lower components are connected. These components are glued together with silicone adhesive. One solution is to add a silicone ring surrounding the part where the upper and lower components are connected. Increasing the actuator's side wall width can also help improve durability.

4 Conclusions

The first actuated soft robotic wearable prototype for assisting and training UE movement in infants was developed and tested on a physical model to evaluate its performance, wearability, and safety. The current design was successful in assisting reaching tasks relatively fast, reliably, and accurately. Further improvements will make this prototype suitable for the next step, which is testing with infants.

The device was designed by taking into consideration the target population characteristics and certain features of existing passive devices that have been found capable of inducing important changes in UE motor function for infants [8,12,13]. Our device is a lightweight garment-based wearable, with integrated low-profile soft components to match the small size and weight of infants' limbs. Maintaining a low profile for the core of the device is important for wearability but also critical regarding our future efforts to make the device more autonomous as more components (e.g., sensors) will need to be added. Preliminary benchmark testing indicates that the current design and placement of the casted soft pneumatic actuators seem appropriate for producing the necessary forces to lift the model's small arms against gravity. Further, the application of the different “muscle” recruitment strategies appears to provide an acceptable and reproducible range of movement reaching trajectories, similar to the movement repertoire seen in infants [15] and a vital component of infant motor learning [16,17].

Nevertheless, there are certain limitations associated with the current design. First, the prototype at this stage does not adapt to changes in real-time. An important component that ongoing work seeks to add in is device sensing. Device sensing can advance both the performance and safety features of the device. For example, the use of electromyography sensors can have the double role of predicting a movement action before it emerges, enabling the actuation of the device to happen at the correct timing, but also may offer information on the work the muscles will be performing so as to estimate fatigue parameters and avoid potential injury in the muscle. However, adding sensors will increase the overall weight which in turn may affect the quantity and quality of limb movement in infants with UE motor impairments whose limited muscle strength has to also overcome their own body mass and gravity [32]. Hence, sensor selection should be addressed carefully. Second, the device at its current form actuates only 2DOF movements per arm. This places a limit on the number of muscle strategies that can be actively supported. Future addition of more actuators will increase the number and complexity of movement strategies we can implement and test, thus leading to additional movement trajectories.

Finally, it is worth noting that increasing the number of actuators used on a wearable device needs to be given careful consideration as it may not always be the right solution. Increasing the DOFs of a device may produce a wider range and more naturalistic movements; however, it also increases the device complexity. The latter tends to increase costs and need for assistance from the user and decrease the safety and number of potential users [33].

Overall, this study introduces the necessary initial building blocks to help introduce a class of novel soft robotic wearable devices to provide assistive feedback to infants with UE mobility challenges.

Acknowledgment

The authors would like to thank Mark Estafanous, David Creighton, Melanie Beltran, Linh Vu, Matthew Sample, Drishti Bhardwaj, and Kristina Rodriguez for their assistance on various aspects of this project.

References

References
1.
Christy
,
J. B.
,
Lobo
,
M. A.
,
Bjornson
,
K.
,
Dusing
,
S. C.
,
Field-Fote
,
E.
,
Gannotti
,
M.
,
Heathcock
,
J. C.
,
O'neil
,
M. E.
, and
Rimmer
,
J. H.
,
2016
, “
Technology for Children With Brain Injury and Motor Disability: Executive Summary From Research Summit iv
,”
Pediatr. Phys. Ther.
,
28
(
4
), pp.
483
489
.10.1097/PEP.0000000000000317
2.
Boyle
,
C. A.
,
Boulet
,
S.
,
Schieve
,
L. A.
,
Cohen
,
R. A.
,
Blumberg
,
S. J.
,
Yeargin-Allsopp
,
M.
,
Visser
,
S.
, and
Kogan
,
M. D.
,
2011
, “
Trends in the Prevalence of Developmental Disabilities in U.S. Children, 1997-2008
,”
Pediatrics
,
127
(
6
), pp.
1034
1042
.10.1542/peds.2010-2989
3.
Maenner
,
M. J.
,
Blumberg
,
S. J.
,
Kogan
,
M. D.
,
Christensen
,
D.
,
Yeargin-Allsopp
,
M.
, and
Schieve
,
L. A.
,
2016
, “
Prevalence of Cerebral Palsy and Intellectual Disability Among Children Identified in Two U.S. National Surveys, 2011–2013
,”
Ann. Epidemiol.
,
26
(
3
), pp.
222
226
.10.1016/j.annepidem.2016.01.001
4.
Lobo
,
M. A.
,
Kokkoni
,
E.
,
Cunha
,
A. B.
, and
Galloway
,
J. C.
,
2015
, “
Infants Born Preterm Demonstrate Impaired Object Exploration Behaviors Throughout Infancy and Toddlerhood
,”
Phys. Ther.
,
95
(
1
), pp.
51
64
.10.2522/ptj.20130584
5.
Cunha
,
A. B.
,
Babik
,
I.
,
Ross
,
S. M.
,
Logan
,
S. W.
,
Galloway
,
J. C.
,
Clary
,
E.
, and
Lobo
,
M. A.
,
2018
, “
Prematurity May Negatively Impact Means-End Problem Solving Across the First Two Years of Life
,”
Res. Dev. Disabil.
,
81
, pp.
24
36
.10.1016/j.ridd.2018.03.007
6.
Lobo
,
M. A.
,
Kokkoni
,
E.
,
De Campos
,
A. C.
, and
Galloway
,
J. C.
,
2014
, “
Not Just Playing around: Infants’ Behaviors With Objects Reflect Ability, Constraints, and Object Properties
,”
 Infant Behav. Dev.,
37
(
3
), pp.
334
351
.10.1016/j.infbeh.2014.05.003
7.
Fox
,
S. E.
,
Levitt
,
P.
, and
Nelson
,
C. A.
,
2010
, “
How the Timing and Quality of Early Experiences Influence the Development of Brain Architecture
,”
Child Dev.
,
81
(
1
), pp.
28
40
.10.1111/j.1467-8624.2009.01380.x
8.
Babik, I., Kokkoni
,
E.
,
Cunha
,
A. B.
,
Galloway
,
J. C.
, and
Rahman
,
T.
,
2016
, “
Feasibility and Effectiveness of a Novel Exoskeleton for an Infant
,”
Pediatr. Phys. Ther.
,
28
(
3
), pp.
338
346
.10.1097/PEP.0000000000000271
9.
Haumont
,
T.
,
Rahman
,
T.
,
Sample
,
W.
,
King
,
M. M.
,
Church
,
C.
,
Henley
,
J.
, and
Jayakumar
,
S.
,
2011
, “
Wilmington Robotic Exoskeleton: A Novel Device to Maintain Arm Improvement in Muscular Disease
,”
J. Pediatr. Orthop.
,
31
(
5
), pp.
44
49
.10.1097/BPO.0b013e31821f50b5
10.
Gemperle
,
F.
,
Kasabach
,
C.
,
Stivoric
,
J.
,
Bauer
,
M.
, and
Martin
,
R.
,
1998
, “
Design for Wearability: Digest of Papers
,”
Second IEEE International Symposium on Wearable Computers
, Pittsburgh, PA, Oct. 19–20, pp.
116
122
. 10.1109/ISWC.1998.729537
11.
Lobo
,
M. A.
,
Koshy
,
J.
,
Hall
,
M. L.
,
Erol
,
O.
,
Cao
,
H.
,
Buckley
,
J. M.
,
Galloway
,
J. C.
, and
Higginson
,
J.
,
2016
, “
Playskin Lift: Development and Initial Testing of an Exoskeletal Garment to Assist Upper Extremity Mobility and Function
,”
Phys. Ther.
,
96
(
3
), pp.
390
399
.10.2522/ptj.20140540
12.
Babik
,
I.
,
Cunha
,
A. B.
,
Moeyaert
,
M.
,
Hall
,
M. L.
,
Paul
,
D. A.
,
Mackley
,
A.
, and
Lobo
,
M. A.
,
2019
, “
Feasibility and Effectiveness of Intervention With the Playskin Lift Exoskeletal Garment for Infants at Risk
,”
Phys. Therapy
,
99
(
6
), pp.
666
676
.10.1093/ptj/pzz035
13.
Babik
,
I.
,
Cunha
,
A. B.
, and
Lobo
,
M. A.
,
2019
, “
Play With Objects in Children With Arthrogryposis: Effects of Intervention With the Playskin Lift Exoskeletal Garment
,”
Am. J. Med. Genet. Part C
,
181
(
3
), pp.
393
403
.10.1002/ajmg.c.31719
14.
LiGreenspan
,
B.
,
Mascitelli
,
B.
,
Raccuglia
,
T.
,
Denner
,
M.
,
Duda
,
K. R.
, and
Lobo
,
M. A.
,
2019
, “
Design of the Playskin Air(TM): A User-Controlled, Soft Pneumatic Exoskeleton
,”
Proceedings of the 2019 Design of Medical Devices Conference
, Minneapolis, MN, Apr. 15–18, pp.
1
4
.10.1115/DMD2019-3231
15.
Konczak
,
J.
, and
Dichgans
,
J.
,
1997
, “
The Development Toward Stereotypic Arm Kinematics During Reaching in the First 3 Years of Life
,”
Exp. Brain Res.
,
117
(
2
), pp.
346
354
.10.1007/s002210050228
16.
Corbetta
,
D.
, and
Thelen
,
E.
,
1996
, “
The Developmental Origins of Bimanual Coordination: A Dynamic Perspective
,”
J. Exp. Psychol.: Hum. Percept. Perform.
,
22
(
2
), pp.
502
522
.10.1037/0096-1523.22.2.502
17.
Fetters
,
L.
,
2010
, “
Perspective on Variability in the Development of Human Action
,”
Phys. Ther.
,
90
(
12
), pp.
1860
1867
.10.2522/ptj.2010090
18.
Chiri
,
A.
,
Cempini
,
M.
,
Member
,
S.
,
Marco
,
S.
,
Rossi
,
M. D.
,
Lenzi
,
T.
,
Giovacchini
,
F.
,
Vitiello
,
N.
,
Chiara
,
M.
, and
Member
,
C.
,
2012
, “
On the Design of Ergonomic Wearable Robotic Devices for Motion Assistance and Rehabilitation
,”
Annual International Conference of the IEEE EMBS
, San Diego, CA, Aug. 28–Sept. 1, pp.
6124
6127
.10.1109/EMBC.2012.6347391
19.
Baniqued
,
P. D. E.
,
Baldovino
,
R. G.
, and
Bugtai
,
N. T.
,
2016
, “
Design Considerations in Manufacturing Cost-Effective Robotic Exoskeletons for Upper Extremity Rehabilitation
,”
Eighth International Conference on Humanoid, Nanotechnology, Information Technology, Communication and Control, Environment and Management,
Cebu City, Philippines, Dec. 9–12, pp.
10
15
.
20.
Roderick
,
S. N.
, and
Carignan
,
C. R.
,
2005
, “
An Approach to Designing Software Safety Systems for Rehabilitation Robots
,”
Proceedings of the Ninth IEEE International Conference on Rehabilitation Robotics
, Chicago, IL, June 28–July 1, pp.
252
257
.10.1109/ICORR.2005.1501096
21.
Lobo
,
M. A.
, and
Galloway
,
J. C.
,
2013
, “
The Onset of Reaching Significantly Impacts How Infants Explore Both Objects and Their Bodies
,”
Infant Behav. Develop.
,
36
(
1
), pp.
14
24
.10.1016/j.infbeh.2012.09.003
22.
Lobo
,
M.
,
Galloway
,
J.
, and
Heathcock
,
J. C.
,
2015
, “
Characterization and Intervention for Upper Extremity Exploration & Reaching Behaviors in Infancy
,”
J. Hand Ther.
,
28
(
2
), pp.
114
125
.10.1016/j.jht.2014.12.003
23.
Rus
,
D.
, and
Tolley
,
M. T.
,
2015
, “
Design, Fabrication and Control of Soft Robots
,”
Nature
,
521
(
7553
), pp.
467
475
.10.1038/nature14543
24.
Rocha
,
N. A. C. F.
,
da Costa
,
C. S. N.
,
Savelsbergh
,
G.
, and
Tudella
,
E.
,
2009
, “
The Effect of Additional Weight Load on Infant Reaching
,”
Infant Behav. Dev.
,
32
(
2
), pp.
234
237
.10.1016/j.infbeh.2008.12.012
25.
Tondu
,
B.
,
2012
, “
Modelling of the Mckibben Artificial Muscle: A Review
,”
J. Intell. Mater. Syst. Struct.
,
23
(
3
), pp.
225
253
.10.1177/1045389X11435435
26.
Shepherd
,
R. F.
,
Ilievski
,
F.
,
Choi
,
W.
,
Morin
,
S. A.
,
Stokes
,
A. A.
,
Mazzeo
,
A. D.
,
Chen
,
X.
,
Wang
,
M.
, and
Whitesides
,
G. M.
,
2011
, “
Multigait Soft Robot
,”
Proc. Natl. Acad. Sci.
,
108
(
51
), pp.
20400
20403
.10.1073/pnas.1116564108
27.
Galloway
,
K.
,
Polygerinos
,
C. P.
,
Walsh
,
C. J.
, and
Wood
,
R. J.
,
2013
, “
Mechanically Programmable Bend Radius for Fiber-Reinforced Soft Actuators
,”
Proceedings of the 16th International Conference on Advanced Robotics
(
ICAR
), Montevideo, Uruguay, Nov. 25–29, pp.
1
6
.10.1109/ICAR.2013.6766586
28.
Mosadegh
,
B.
,
Polygerinos
,
P.
,
Keplinger
,
C.
,
Wennstedt
,
S.
,
Shepherd
,
R. F.
,
Gupta
,
U.
,
Shim
,
J.
,
Bertoldi
,
K.
,
Walsh
,
C. J.
, and
Whitesides
,
G. M.
,
2014
, “
Pneumatic Networks for Soft Robotics That Actuate Rapidly
,”
Adv. Funct. Mater.
,
24
(
15
), pp.
2163
2170
.10.1002/adfm.201303288
29.
Roche
,
E. T.
,
Wohlfarth
,
R.
,
Overvelde
,
J. T.
,
Vasilyev
,
N. V.
,
Pigula
,
F. A.
,
Mooney
,
D. J.
,
Bertoldi
,
K.
, and
Walsh
,
C. J.
,
2014
, “
A Bioinspired Soft Actuated Material
,”
Adv. Mater.
,
26
(
8
), pp.
1200
1206
.10.1002/adma.201304018
30.
Polygerinos
,
P.
,
Wang
,
Z.
,
Overvelde
,
J. T.
,
Galloway
,
K. C.
,
Wood
,
R. J.
,
Bertoldi
,
K.
, and
Walsh
,
C. J.
,
2015
, “
Modeling of Soft Fiber-Reinforced Bending Actuators
,”
IEEE Trans. Rob.
,
31
(
3
), pp.
778
789
.10.1109/TRO.2015.2428504
31.
McDowell
,
M. A.
,
Fryar
,
C. D.
,
Ogden
,
C. L.
, and
Flegal
,
K. M.
,
2008
, “
Anthropometric Reference Data for Children and Adults: United States, 2003–2006
,”
Natl. Health Stat. Rep.
,
22
(
10
), pp.
1
48
.http://ghk.h-cdn.co/assets/cm/15/11/550017f045e74_-_nhsr010.pdf
32.
Groenen
,
A. A. T. K.
,
Kruijsen
,
A. J. A.
,
Mulvey
,
G. M.
, and
Ulrich
,
B. D.
,
2010
, “
Constraints on Early Movement: Tykes, Togs, and Technology
,”
Infant Behav. Develop.
,
33
(
1
), pp.
16
22
.10.1016/j.infbeh.2009.10.002
33.
Reinkensmeyer
,
D. J.
, and
Boninger
,
M. L.
,
2012
, “
Technologies and Combination Therapies for Enhancing Movement Training for People With a Disability
,”
J. NeuroEng. Rehabil.
,
9
(
1
), pp.
17
10
.10.1186/1743-0003-9-17