A Compact and Portable Exoskeleton for Shoulder and Elbow Assistance for Workers and Prospective Use in Space

Exoskeletons are wearable robotic devices that surround the anatomy of the user to work in tandem with it. Depending on their structure, exoskeletons can be classified as rigid or flexible. The structure of flexible exoskeletons is made of soft materials, such as fabrics, which adapt to user motion. Therefore, these devices are prone to becoming misaligned with the user, due to improper fitting and slipping of the exoskeleton components on the user body. This article describes a cable-driven exosuit, called LUXBIT, that favors its anatomical adaption to the user by arranging the fabric fibers and sewing patterns to transfer the mobilizing forces. This prototype integrates a novel deformable mechanism that promotes the natural lifting of the arm. LUXBIT is intended for bimanual assistance in daily living, being equipped with a backpack to this end. The results analyzed in this article show that LUXBIT reduces muscle activity in the upper limbs’ flexion by ratios of up to 13.17% and allows the user to hold tiring postures for 62.91% longer.


I. INTRODUCTION
H UMAN-ROBOT interaction (HRI) is a field that has been progressively growing as robots move into society. Their continuous integration into everyday tasks involves a set of challenges for effective collaboration and cooperation between humans and robots [1]- [3]. Exoskeletons provide, in this context, an excellent example of a novel human-robot collaborative environment.
Exoskeletons are essentially robots positioned over the human anatomy to assist, oppose, or augment human locomotion. Their benefits in medical and working environments have boosted their use over recent decades. Exoskeletons can be classified attending, depending on their structure, as rigid or flexible [4]- [7].
Briefly, rigid exoskeletons have a solid self-supportive structure that can autonomously operate while delivering high assistance forces. Thus, rigid exoskeletons are extensively used in muscle restoration and heavy-duty tasks. Nonetheless, the rigid linkages that these devices position between the user's articulations can become restrictive for daily living gestures.
In contrast, flexible or soft exoskeletons have a light structure that adapts to user motion. These flexible exoskeletons transfer mobilizing forces to different points of the human anatomy in such a way that drives locomotion [8]. In this way, their operation is supported by the wearer's anatomy and is hence highly dependent on the user's motion intention. Moreover, such functioning exposes soft exoskeletons to becoming misaligned with the user, which eventually leads to under-assistance. Over the years, various works have proposed tightly fitting these soft exoskeletons to the user, with most being based on fastening fabrics and components to the user. Another proposal commonly adopted in cable-driven systems consists of twisting the cables around the limbs.
Turning to their application, flexible exoskeletons excel in metabolic cost reduction and fatigue delay [9]. Their versatility and nonrestriction of the natural user pace have led to them being proposed for daily living and outdoor activities [10]- [13].
However, soft exoskeleton solutions have not been deployed in the literature to the same degree for lower and upper limbs [14], [15]. The redundant kinematics of the upper limbs and their wide workspace might be one of the possible reasons. Specifically, the mobility of the arms entails a significant surface deformation of the anatomy, which can expose the robot to becoming misaligned with the user [16]. Therefore, the successful deployment of exosuits for the upper limbs requires from new sensing and actuation strategies that tackle the anatomical adaption of the device [17]- [19]. One of the interests in advancing soft exoskeletons relates to the prevention of musculoskeletal conditions (MSCs). The World Health Organisation (WHO) has reiterated the urgency of developing new tooling and equipment to mitigate the growing incidence of these disorders of daily living [20], [21]. Similar conditions also appear in weightlessness situations like those in space applications, such as the International Space Station (ISS), and the future Lunar Space Station (Gateway), which seek methods to palliate the effects of microgravity in astronauts [22]- [25]. Thus far, the literature has leveraged physical training to deal with muscle atrophy and bone tissue in space missions. [26]. However, prospective space plans include a set of new weightlessness conditions and long-term missions, such as those in NASA's Artemis Base Camp, Mars Base Camp, and NextStep [27]- [30], for which the National Space Biomedical Research Institute (NSBRI) is seeking solutions and preventive equipment [31]- [33].
In this context, the lightweight and compactness of soft exoskeletons make them suitable assistive devices to be deployed in space missions, where they can also enhance astronauts' performance by reducing the loss of dexterity and proprioception that occurs in space [26], [34], [35]. Specifically, the concept of dual-mode exoskeletons is being applied to exoskeletons that can both assist the operator to perform tasks but also work in reverse to increase muscle activity [36]- [38].
This article describes and evaluates the new portable prototype of the LUXBIT exoskeleton. LUXBIT [39], [40] shares a number of components with EXOFLEX [41], which is a flexible exoskeleton intended for rehabilitation. LUXBIT is entirely based on textiles and deformable components that deliver the open-loop cable-driven transmission explained in Section II [42]. In our prototype, we have exploited the tailoringbased concept to minimize the rigid components and weight on the user. Specifically, it is equipped with a new mobile pulleybased system for flexing the elbow and a novel mechanism to lift the shoulder. The elbow mechanism applies the principle already shown in [40] but has a new and more robust embedding method by using circular dovetails to fast the parts riveted to the button's sides. The redesigned shoulder actuation enhances the adaptation of the exoskeleton to discrepant anatomies and promotes the natural movement of the arm. This new solution also reduces the number of actuators in the system. To evaluate the system, nine participants were asked to use the exoskeleton as described in Section III. The results in Section IV show that the exosuit can reduce muscle activity by up to 16.63% and extend the time an individual can remain in a tiring pose by up to 43.95%. Table VII shows a summary of the benefits of this new prototype.

II. DESCRIPTION OF THE LUXBIT PORTABLE EXOSUIT
The LUXBIT exoskeleton assists elbow flexion entirely and can lift the shoulder up to 120 • using two motors placed in a backpack. Fig. 1 shows the construction, which uses stainless steel ropes in 7 × 7 and 1 × 19 arrangements with diameters of 0.75 and 1 mm. The minimum breaking load for the elbow and shoulder tendon is 400 and 520 N, respectively. These cables Optitrack markers mounted on a complementary fixation (dark piece) to ensure tracking of shoulder pivot. The image also shows a prototype to extend the elbow (posterior forearm and grey clamp), which is not discussed in the article. The arrows mark the cable pulling directions. A color was assigned to each articulation to simplify identification (red for the shoulder and green for the elbow). The shoulder pivot has a passive joint that aligns it with the arm. are guided by PTFE sheaths and have a thin nylon coating that prevents puncturing, dismantling, and grease leaks. The LUXBIT components in the figure are deformable and have a button-based attachment to anchor them to the base garment.
LUXBIT is based on textiles, in the form of various textile patches, such as bracelets, armbands, and shoulder pads designed to embed the actuation components and enhance the anatomical adaption of the commercial base garment. The design of these textile add-ons consists of properly directing the warp and weft fibers so as to promote sectional limb adaption and dampen the forces leading to misalignment. This arrangement of textiles is shown in Fig. 2. Fig. 2 also displays the commercial textile backpack that makes the exoskeleton portable. The backpack is ergonomically certified to carry payloads up to 20 kg. Three modifications were applied to adapt it to the exoskeleton. First, the stuffing in the deltopectoral area of the handles was removed so that the pivoting elements better fit the user. Second, a foamed mesh layer was added to the user back with an elastic connection to the wristband that improves comfort. Finally, the third enhancement consisted of adding elastic straps over the handles that wrap around the user trunk. In this way, the total weight of the backpack for both arms is 1.153 kg, whereas each arm is only loaded with 265 gr. Bolted camping buttons were used to attach the actuation components to the backpack and prevent them from becoming detached during operation. Regarding the actuation, the shoulder is assisted with a 24-VDC motor with a gearbox and encoder Maxon DCX22S+GPX26HP+ENX16EASY. In contrast, elbow assistance is delivered by a linear Nema-14 with a 140-mm rod. The control uses a previously described super-twisting sliding mode controller [41]. The model integrates a formulation that considers various anatomical parameters of the user, such as the section area or length of the arm segments [40].
In terms of power, the exosuit delivers up to 2.5-kg assistance, while raising the arm in five seconds and flexing the elbow in three seconds. This power limitation was chosen to prevent anatomical overloading and reduce the forces transferred to the user over the deltopectoral area. Moreover, the detachment resistance and pivoting elements were designed so that they detach or break when the tendon tension surpasses a maximum safety value.

A. Textile-Coupling Interface
The physical interface to transfer the mobilizing force to the user is an essential element for the performance of the exoskeleton. The coupling trades the comfortability of the solution and the quality of the fastening to prevent perturbations, such as slipping, which lead to under-assistance. Additionally, it must consider the effects of nonassisted gestures. A widely used coupling method for cable-driven exoskeletons consists of embedding the cable anchor into a garment by a sewn, thermal, or chemical link. In this way, the cable pulls the anchor, which transfers the mobilizing force and levers the limb. An important element in cable-driven exoskeletons is the cable anchor, which is commonly fastened to the limb that the tendon pulls. Thus, the fabric needs to tightly fit the limb or the garment would be prone to slipping. An elastic component is also introduced into the actuation when the anchor is not tightly fitted.
The literature shows that these solutions are commonly fitted to the user by Velcro straps, belts, and thick, compressive fabrics (such as neoprene) [43], [44]. The main disadvantage of this method is that it can lead to skin conditions such as reddening, irritation, and numbing. Similarly, nonbreathable fabrics can cause condensation and humidity. Fig. 3. On the left, the behavior of the textile clamp when the cable is pulled. On the right, the elbow mechanism. The pulley-in-button has a self-locking mechanism to prevent its detaching. The cable pulls in the direction of the arrows.
The LUXBIT exoskeleton uses a textile-based coupling that promotes comfort and anatomical adaptation. Moreover, it only presses the skin when the exoskeleton is assisting the user. Fig. 3 shows this v-shaped textile clamp where two layers of ProCool Athletic Interlock Silver fabric with SILVADUR are stacked together. The relative orientation of each layer arranges the stiff fibers, the warp, so they go from the clamp's symmetry axis to one of the button faces riveted to the clamp extremes. In this way, the pulling of the tensed cable is propagated through the different stitches toward the symmetry axis of the clamp, where an effective force is created that levers the limb. The shape of the stitches and clamp follows a previously published 3-D-CAD procedure [40]. However, in this prototype we have redesigned the pivots and components that pull the limb to facilitate their assembly. The shoulder textile clamp has also been adapted to the novel shoulder mechanism by recalculating the shape for the cable angle and considering its fixation at an intermediate position between shoulder flexion and abduction (45 • of horizontal flexion). Moreover, the stitches are padded by 1-cm to avoid their sinking into the skin.
The clamps are connected to an elastic armband of ProCool Stretch-FIT Dri-QWick with a straight stitch over the symmetry axis. The armband increases the friction with the limb so it helps to prevent misalignment. Note that the fabrics in both the armband and clamp are thin, breathable, and antimicrobial.

B. Cable Routing
The cable routing and anchoring solution described in this section favor a lightweight, compact, and anatomically adaptable transmission, compatible with nonassisted gestures (pronationsupination and horizontal flexion). Cables have been routed so their path is not altered by the high anatomical deformation that the body undergoes during locomotion, such as muscle contraction or joint displacement. In this way, the elbow flexion cable positions two mirrored pivot points on the sides of the arm. The pivots consists of buttons whose top face has been riveted with a custom element to anchor a 120 • -curved sheath. After assembly, the sheath is anchored with screws.
Similarly, the button on the forearm clamp has a 160 • -curved guide embedded in it. It also integrates a circular dovetail that is fastened by rotating the top part after both button faces have been buttoned up. In this way, the routing for elbow flexion starts at the triceps, where the cable is anchored by a 3-D-printed custom buckle. To this end, the buckle has a slot where a cable lace can be anchored by a screw (see Fig. 3). The cable is guided to the pivot on the inner sidearm towards the textile clamp over the forearm. Then, the curved guide on the textile coupling redirects the cable towards the pivot on the outer sidearm, where a sheath guides it to the motors in the backpack. A new complementary guide has been positioned over the posterior-middle deltoid to maintain the cable over the arm even during scapular displacements. This guide has also been embedded on a button.
This routing by the sides of the arm reduces the perturbations from the biceps contraction. Additionally, there are redundant collocations for the pivot on the armband so as to promote anatomical adaptation. Another advantage of routing the cable in the same way as in a mobile pulley systems is that it makes the system compatible with glenohumeral rotation. Moreover, the lateral pivots create a zero net lateral force, unlike previous prototypes [40].
For the shoulder, the scheme is simpler: the cable is anchored to the button on the textile clamp and a rotational pivot lies over the shoulder. A sheath was embedded in the shoulder pivot to guide the cable to the motors. In our previous designs, we used a pierced button to anchor the cable. However, this solution involved issues in terms of wearability and readiness: the cable lace avoided the clamp being completely open without dismantling the whole actuation, nor could maintenance be carried out without loosening both ends of the cable (such as relacing or shortening the cable) [39], [40]. Our prototype uses a novel solution with two custom components that are riveted to each of the faces of the button. These small 3-D-printed parts integrate a dovetail mechanism to increase the fastening resistance, and one of them integrates has a slot where the cable lace can be fastened with a screw (see Fig. 4).
Similarly, in our previous publications, we discussed the issues detected using independent pivots. Briefly, mechanisms placed over the shoulder surroundings to assist a pure motion, such as the abduction or flexion of the shoulder, are difficult to fit to discrepant anatomies. The human shoulder surroundings are highly movable and depend on a plethora of genetic factors. A significant point of concern was that each user tended to laterally displace the arm during shoulder flexion. In this way, the resulting anatomical under-fitting created lateral forces on the component during operation that made it press the skin or under-assist the shoulder.
In this version of LUXBIT, we propose a new pivoting mechanism has been proposed that allows the cable to be realigned with the arm and, hence, fit discrepant anatomies. Additionally, it reduces the net assistance force required and better distributes the forces over the user.

C. Rotational and Deformable Shoulder Pivot
The design of the new shoulder pivot promotes anatomical compatibility and noninvasive assistance by allowing the mechanism to realign during assistance. The mechanism in Fig. 4 consists of a deformable pivot with a passive joint that rotates around a grooved base lying over the trapezius. The grooves in this wide base allow it to deform and thus fit the trapezius geometry of the user by using four buttons, one on each corner of the base, and an elastic strap. Two halves of the elastic strap were sewn to the shoulder-pad underneath the base. Each of them goes inside one of the slots in the base and then turns around the trunk of the user, fitting the base to the trapezius surroundings (see Fig. 4). There is an additional hidden button under the pivot to prevent reaction forces from curving the base. The shoulder pad is derived from the one described in previous prototypes, although this new model uses thinner fabrics (Dry-Qwick) combined with reinforcement mesh, integrating various elastic areas to ensure the fit. Additionally, its stretchable connection with the armband under the shoulder clamp was modified to twist around the arm.
The shape of the pivot was obtained by applying compliant mechanism theory, using a double curvature that adds support points over the shoulder as the piece deforms. In this way, the pivot gradually stands on the deltoid muscle. Despite this deformation, the pivot height can be maintained constant during gestures. This construction allows the pivot to be forwarded 6.18 cm from the shoulder, which halves the force requirement from our previous solutions as well as mitigates undesired force components that push the humerus into the articulation [42].
The pivot is mounted over the grooved base by using a passive joint that allows the pivot to freely rotate around the shoulder. To this end, three slots were integrated into the pivot shape: two of these anchor the guiding sheath, while the frontal one is directly pressed by the cable when the pivot is unaligned with the arm. Such pushing of the cable forces the pivot to rotate until properly aligned with the arm. The multimedia shows how the pivot dynamically adapts to assist consecutive flexion and abduction.

III. METHODS AND DATA PROCESSING
Our evaluation seeks to estimate the benefits of an exoskeleton when the participants move naturally. The study accepts the following assumptions.
1) The speed of voluntary and daily gestures is lower than the reactive muscle response.
2) The proposed speed can be surpassed with the only drawback of not perceiving the assistance of the device. 3) The exoskeleton does not provide sufficient force to completely mobilize the limbs, especially the shoulder, requiring the user to be actively engaged. 4) Cooperative situations require the user muscle to be active, and thus, the muscle benefit is less apparent than when the user lets the device guide its motion; positive outcomes are yet to be detected. This evaluation is based on nine healthy and well-informed participants aged 25.55 years ±4.15, who wore the device and performed five repetitions of elbow flexion, shoulder abduction, and shoulder flexion. The ethical guidelines of the Universidad Politécnica de Madrid (UPM) were followed. The participants had no prior experience of the use of exoskeletons, but were allowed to test the prototype before the experiment. All the gestures started with the participant standing, and the arm relaxed next to the trunk. The task was performed first unassisted, then assisted, and was finally repeated holding a 1-kg dumbbell.
The motion pattern performed was as follows: flexing the articulation up to 90 degrees, holding this position for 2 s, extending the articulation, and remaining in the lower position for another 2 s. The elbow flexion had to be accomplished in 4 s and the shoulder in 6 s. This difference was due to the maximum speeds limitations of the actuators. In this case, the participants did not receive postural feedback but were trained on the gestures and the device. The participants rested for at least 5 min after the training and 1 min between trials.
Regarding data acquisition and processing, an NVidia Jetson Nano with three MyoWare muscle sensors measured the sEMG signal of the biceps, trapezius, and deltoid muscles during the tests. Meanwhile, OptiTrack monitored their motion by using nine markers: three over the forearm, arm, and the pivot piece of the shoulder. The collocation is a modified Biomech to avoid occlusion [45]. The data were processed offline in Motive and MATLAB. The misfit data from Optitrack were manually labeled with Motive, and the surface electromyography (sEMG) signals were normalized using the maximum activation measured in each participant. The maximum voluntary contraction (MVC) test was performed after the experiments and consisted of two phases for each muscle measured. In the first, participants pulled from a 10-kg elastic band to perform isometric contraction of the muscle measured. A 90 • flexion (or abduction) was maintained until muscle exhaustion was felt. Subsequently, the participants were asked to perform 20 repetitions of the corresponding gesture in the experiment but, this time, pulling the elastic band.

IV. RESULTS AND DISCUSSION
The following sections analyze the results obtained for each of the experiments. They are summarized in two tables for each of the gestures studied. One of these tables shows the normalized muscle activity and trajectory, while the other expresses the percentage of variation generated by holding a 1 kg weight when the exoskeleton was active and nonactive. The purpose of this table is to detect how the exoskeleton modifies the impact of grasping a weight; a desirable outcome would be the device reducing the impact of picking up a weight in trajectory performance and muscle activity. Table I and Table II shows a significant mean reduction of 9.63% and 9.45% in metabolic cost when using the exosuit, for the loaded and loaded case, respectively, with two-tailed Wilcoxon sign test p-values ≤0.05 with a 0.05% confidence interval. The Shapiro-Wilks and Fisher tests indicate that the distributions cannot be assumed as normal at p level ≤0.03 but similar in variance with p-values ≥0.95. Moreover, the onetailed Wilcoxon test strongly supports the alternative hypothesis that the exosuit reduces muscle activity at p-values ≤ 0.01.

A. Elbow Flexion
In contrast, the exoskeleton shows no significant effect on the sEMG peaks (paired two-tailed t-test with p-values of 0.3228 and 0.8386, under Shapiro and Fisher p-values ≥0.08 for the unloaded and loaded cases). However, there are cases where the power spike occurs at the maximum point of the gesture, indicating that the device is not correctly assisting the last part of the gesture, as in Fig. 5(a). Indeed, the trajectory shows that the upper position held by the participant, rather than being constant, creates a slope.
Such spiking of the sEMG and variability in the upper pose when using the exoskeleton may be related to friction between the Optitrack markers and the tendons. An insufficient rod length of the linear motors is another possible reason observed for the tallest participants (above 1.80 m), who required the full rod length. A third possible reason for the upper angle's underassistance relates to the users who moved their arm upward beyond the reference position, as in Fig. 6. This last event demonstrates that the exoskeleton allows the free movement of the user, who can continue the movement despite the robot having stopped.
The trajectory results show a similar behavior in users that held the peak position for longer than the proposed 2 s. Nonetheless, the holding of the upper flexion notably improved by an average of 41.19% and 49.16%, with and without load, when the exoskeleton was active with Wilcoxon sign p-values ≤0.01 (Shapiro p-values ≥0.08 and Fisher p-values ≤0.05). Another notable aspect of the trajectory in both graphs is that participants did not extend the elbow completely when the exoskeleton was inactive. Such gradual alteration of the range of motion has previously been reported as a sign of muscle fatigue [46]- [48], although proprioception and attention loss might also be considered, even given the short duration of the experiment. However, it is worth noting that the exoskeleton, despite not actively assisting elbow extension during the experiment, achieved a lower and less variable resting angle. In this context, the exosuit improved the users' trajectory and proprioception. The participant in Fig. 6, for example, erratically followed the time reference proposed when the device was inactive.
sEMG data was normalized for each participant using a maximum effort test. P is for peak. Duration refers to the total duration of the test, and hold time is the average time that participants maintained the upper position.
The table shows the relative impact of grasping a weight in different parameters. These values are obtained as the difference divided per the value without weight. A negative value indicates the percentage by which the weight reduced the parameter, whereas a positive means that grasping weight increases the parameter.

B. Shoulder Flexion
The results of the exoskeleton for shoulder flexion are apparently moderate regarding muscle activity (see Tables III and  IV). Specifically, the reduction of muscle activity by 2.75% and 2.26% for the loaded and unloaded cases is not statistically significant according to the p-values of 0.3125 and 0.8695 delivered by a two-tailed Wilcoxon sign test. Similarly, the reduction of the sEMG peaks by 11.92% and 4.73% cannot be considered important at a significance level of 0.05 (p-values ≥ 0.5993).
Nonetheless, these results are impressive when considering that participants reached a higher flexion angle and maintained the upper flexion 36.90% and 43.95% longer while using the exoskeleton, under a one-tailed Wilcoxon sign with p-values ≤0.01. Furthermore, the weight was shown to not vary the assistance perceived (two-tailed p-values ≥0.2031). The exoskeleton can thus cut down the muscle stress to which the user is exposed during demanding gestures. The blue-shaded convex patterns in the trajectories of Fig. 7 are another interesting event that denotes a change in speed during the gesture. The two possible interpretations in this regard relate to frictions that might have hindered the speed of the exoskeleton and the users that might have exceeded the proposed speed and thus detached assistance. Two elements of evidence support the latter. First, Fig. 7 illustrates how participants performed a similar convex trajectory in nonassisted gestures. Second, participants had no experience or training in human-robot interaction with wearable exoskeletons. This lack of proper HRI knowledge can impair the assistance obtained by the device, especially when considering that the exosuit is not sufficiently powerful to completely assist the shoulder.
The exoskeleton does not alter the variability of the upper and lower angles whose one-tailed Wilcoxon sign tests delivered p-values ≥ 0.5469, meaning that such variability can be considered intrinsic to the participant's proprioception. The trajectory of the second participant in Fig. 7 reveals another interesting  III  AVERAGED ASSISTANCE FOR SHOULDER FLEXION   TABLE IV  IMPACT OF WEIGHT ON THE SEMG AND TRAJECTORY WHEN THE USER FLEXES THE  case: the exoskeleton allows the shoulder to be accommodated when holding the upper flexion. Specifically, most repetitions include an over-flexion of the shoulder from 5 • to 10 • , followed by a descending motion after the holding position. This can be interpreted as an anatomical rearrangement to accommodate the shoulder girdle and internal shoulder joints. This event also takes place at the lower positions, where participants could modify the resting pose freely. sEMG data were normalized for each participant using a maximum effort test. P is for peak. Duration refers to the total duration of the test, and hold time is the average time that participants maintained the upper position.
The table shows the relative impact of grasping a weight in different parameters. These values are obtained as the difference divided per the value without weight. A negative value indicates the percentage by which the weight reduces the parameter, whereas a positive means that grasping weight increased the parameter.

C. Shoulder Abduction
Shoulder abduction outcomes, in Tables V and VI, are similar to those obtained for shoulder flexion: the exoskeleton makes it possible to hold tiring poses for 33.73% and 36.92% longer, without generating significant differences in the deltoid muscle's sEMG. Specifically, the impact on the holding time is significant at one-tailed p-values ≤ of 0.01, whereas p-levels ≥0.2969 underpin the nonvariation of the sEMG in two-tailed paired tests. Surprisingly, the exoskeleton was found to reduce the muscle activity of the trapezius muscle by 16.63% in the loaded case, with a significance of p-level ≤ 0.05 in a one-tailed Wilcoxon sign test. Another two-tailed Wilcoxon test also shows that the differences in the peak muscle activity in Tables V and VII are nonsignificant (p value ≥ s0.2958).
Another key result is the increase in the maximum angle reached in the unloaded case, which, additionally, does not entail an increase in muscle activity. However, there is a notable difference in the upper angle between the unloaded and loaded cases. One of the most important attributable causes for this specific case is the backpack handle's displacement due to  V  AVERAGED ASSISTANCE FOR SHOULDER ABDUCTION   TABLE VI  IMPACT OF WEIGHT ON THE SEMG AND TRAJECTORY WHEN THE USER ABDUCTS THE SHOULDER   TABLE VII  AVERAGED RELATIVE ASSISTANCE ACCOMPLISHED BY THE EXOSUIT, COMPARED TO THE UNASSISTED CASE improper adaptation to the participants. Participants with small scapular distances (EU S size) required the addition of a 1-cm foam so as to prevent the width of the pivot base from pressing the upper deltoid during the gesture. In these cases, the observed lateral sliding of the handle may have reduced the assistance delivered. Another explanation relates to participants stopping the movement early when the shoulder pivot touched their deltoid muscle, despite this behavior being expected, notified, and trained before the experiments.
Regarding trajectory, although the tables and Fig. 8 show the peak angle to be significantly varied by the exoskeleton, this difference is not statistically significant (p-values ≥ 0.42). Thus, it can be attributed to the user proprioception, given that this effect is reduced by load. The impact of the weight in assistance is likewise unnoteworthy. sEMG data were normalized for each participant using a maximum effort test. P is for peak. Duration refers to the total duration of the test, and hold time is the average time that participants maintained the upper position.
The table shows the relative impact of grasping a weight in different parameters. These values are obtained as the difference divided per the value without weight. A negative value indicates the percentage by which the weight reduces the parameter, whereas a positive means that grasping weight increased the parameter.
A positive value denotes the percentage by which the exosuit improved a parameter. A negative value indicates that the device had a negative effect on the parameter. The percentage refers to the unassisted value.
V. CONCLUSION This article describes the mechanisms for assisting elbow flexion and shoulder uplifting in the LUXBIT exoskeleton. The exosuit uses a textile backpack to promote outdoor application and freedom of movement around the workspace. The article shows that fabric-layer stacking and tailoring methods deliver a lightweight and compact actuation that properly withstand the actuation forces. The novel concept of embedding the actuation components to be riveted with buttons resulted in a detachable system that favors the wearing and maintenance of the exoskeleton. Compared to our previous version using stacked-layer armbands [40], the redesigned components with circular dovetails ease the self-wearing of the device and reduce preparation time.
Previous designs of LUXBIT required the cable to be maintained inside the pulley for attaching the elbow mechanism. This task is simplified by the new solution where the cable is integrated using the PTFE curved sheath in the button. A similar approach facilitates the detaching of the pivots and shoulder anchoring. Additionally, the new anchoring methods allow the cables to be tightened, and their length adapted to the user anatomy. Despite the self-tightening features in the controller, our previous works revealed the convenience of using a shorter/larger cable for significant height differences between subjects. This process was notably tedious for the shoulder as it required dismantling the exoskeleton. In contrast, the new anchoring and guiding components facilitate it. Moreover, adding a circular dovetail has allowed the robustness and anchoring force of the buttons to be configured.
Another important outcome was found in the anatomical adaptation of the new shoulder pivot with one-degree of freedom. Its grooved base was designed to fit discrepant anatomies by fastening an elastic belt. Similarly, the addition of a passive joint to assemble this base with the pivot allowed the mechanism to be dynamically realigned with the arm during the gesture. Contrarily, our previous prototypes tried to tackle each gesture with separate mechanisms, which revealed the exoskeleton to be sensitive to the user's anatomy [39], [40].
Specifically, [39] shows the lateral forces created by such anatomical discrepancies while evaluating shoulder flexion. In contrast, the new rotational and deformable solution has benefited both the natural pace of the arm and the weight of the system by reducing the number of motors required. Moreover, this novel prototype is the first to our knowledge to successfully evaluate assisting the abduction gesture with flexible textile-based methods in a statistically relevant manner.
The results from nine participants confirmed the effectiveness of the exosuit in assisting upper limb gestures. Specifically, the device was proved to reduce muscle activity by up to a 9.63% and 3.99% for the biceps and deltoid, respectively, while extending the time that users could hold a tiring position by 49.61%. These results are noteworthy when considering that holding such poses entails a higher muscle stress, which the user, thanks to the exoskeleton, does not suffer. The most striking results in this context were delivered by the trapezius muscle, whose activity during the abduction gesture was reduced by up to a 16.63%. Table VII shows the most important results in this regard. Additionally, the device improved the linearity and repeatability of the gestures without anatomically obstructing accommodative movements. Indeed, the coupling clamp proposed was shown to allow the lower and upper angles held to be modified in a range of 5 • .
Consequently, the assessment of the exosuit shows the suitability of the device for repetitive and maintained postures such as those performed in manual work (overhead operation, hammering, holding a weight, screwing-unscrewing in uncomfortable postures, etc.). The results demonstrate the potential application of the device to the prevention of musculoskeletal conditions. In this context, signs of fatigue in participants, such as gradually reduced elbow extension, were reduced by the device and the overall range of motion improved. These results likewise suggest that the device can correct improper postures derived from attention loss or distraction, although this statement requires proper evaluation to be accepted. Moreover, the effects of microgravity in astronauts have been widely reported. The muscle mass decline and bone mass loss generated in these situations have been shown to trigger various health conditions, such as osteoporosis, that can be counteracted by physical training and, especially, by high intensity exercises. However, new tooling and methods that can be taken on board are in continuous research. Lower limb dual-mode control exoskeletons have recently been proposed as novel devices that aid astronauts through the workday but also oppose their locomotion in safe environments as a countermeasure for microgravity. An upper limb and portable exosuit are therefore likely applicable to relieve the muscle stress of demanding tasks and prevent unsafe movements in all these situations. Moreover, it is the authors' opinion that exoskeletons can likewise improve the performance of astronauts by mitigating the effects of weightlessness conditions in the upper limbs' proprioception, dexterity, and force exertion. Thus, our prospective schedule includes the evaluation of the long-term effect of the exoskeleton in occupational bimanual tasks and approaching its dual-mode functioning.

VI. LIMITATIONS OF THE STUDY
Three limitations can be highlighted in our study conducted. Assistance power was limited to ensure all participants could bear it. This can be increased by removing the software limitations or modifying the motors. In this context, to deliver assistance of over 3 Kg, it is advisable to add reinforcement layers to the printed components and position two plastic disks during the riveting of the buttons (see [42]). Similarly, the muscle activity of the triceps muscle could have provided further information on the impact of the exoskeleton on the users. Finally, measuring the forces transferred by the textile clamps and armbands is a prospective improvement that might improve the human-robot interaction of the system. As a final note, we would like to highlight that these results cannot be directly correlated and compared with the ones in our previous publications due to the different basis assumed. In this article, the users were actively engaged, and thus, the benefits of the exoskeleton, especially for muscle activity, significantly differ from those measured when the user is entirely mobilized by the device (as in our previous evaluations).