The study design was a randomized crossover study with repeated measurements (work activities with and without wearing the exoskeleton, before and after fatigue, under single- and dual-task conditions) under laboratory conditions (cf. Figure 1).

A total of 10 healthy participants (25.3 ± 2.5 years; 174.6 ± 10.2 cm; five female) were recruited from members of the University of Hamburg (employees and students) (cf. Table 1).

Overall, the participant’s task (single task [ST]) was to apply nails overhead with a nail gun in three predetermined difficulty levels based on Fitts’ law (further referred to as Fitts’ tasks) as quickly and accurately as possible. Depending on the randomization plan for the order of conditions, they performed the nailing task with and without an upper-body exoskeleton, as well as with and without previous physical fatigue based on the fatigue protocol. Within each of the four conditions, the nailing task was conducted with and without the additional secondary task in a randomized order. The secondary task was to audibly count all the beats of a metronome set at a frequency varying between 50 and 70 bpm throughout the entire task. Overall, the precision errors during nailing and the errors in counting the sound signals were recorded under both conditions (cf. Figure 2).

The Fitts’ task is a standard motor control experiment in which participants move quickly between two targets of different sizes and distances (29). The aim is to assess the speed and accuracy of the target movements. The difficulty of the task is determined by a difficulty index, which is calculated based on the size and distance of the target areas. In this study, each stencil contained four targets (squares) at each corner. The size of each square and distance between squares have been set according to Fitts’ law equation to determine three indexes of difficulty: ID = log2 (2D/W) (ID: index of difficulty, D: distance between targets, W: width of each target). The three IDs have been set to increase linearly in difficulty (ID1 = 2.3; ID2 = 3.9; ID3 = 5.4).

The participants fulfilled the task in a standing position under a wooden board, equipped with the paper-based predetermined Fitts’ task patterns, into which they were supposed to place the nails. The height of the board was determined based on the condition that the elbows reached shoulder height as soon as the nail gun was in contact with the wooden surface. At a signal from the instructor, the participants began the task while the time measurement began (cf. Figure 2).

In the DT condition participants were asked to count the number of tones provided by a metronome with varying frequency around 50–70 bpm.

Participants wore a passive upper-body exoskeleton during the task, which uses the tension of elastic elements when the arm is lowered to provide the necessary energy for mechanical support when the arm is raised. The exoskeleton, which was used in this study (Skelex 360), was put on by a person according to the instructions, like a backpack. It was then adjusted based on anthropometric measurements indicated by numerical values. The aim is to position the support component on the upper arm as close as possible to the elbow without exceeding it. A uniform support force was selected for all participants and tasks. No further individual settings could be carried out by the participants (cf. Figure 2).

Following previous research (11, 12), we decided to assess muscle activity of the shoulder as well as from the calf muscles by performing weight exercises in three sets. Each set consisted of a dynamic 2-min calf raise and a dynamic 2-min front raise exercise with weights. The men used a weight of 5 kg, the women a weight of 2.5 kg. A rest period of 60 s was observed between the individual sets. During the execution of the exercises, a metronome was set to a tempo of 60 beats per minute (bpm), and both the front lift and the calf raise were to be performed at this tempo (cf. Figure 2).

The primary outcomes of the study were muscle activity and the time taken to place a nail into a square in each Fitts’ task, as well as the error rates in the DT scenario for the auditory counting task. The main focus was on the difference induced by the use of the upper-body exoskeleton as well as by previous fatigue.

The following statistical analysis were conducted in a fully blinded procedure with references to the research questions.

However, it should be noted that possible confounding factors, such as previous experience with exoskeletons or individual physical condition, were not included in the statistical analysis. However, this was asked in advance and all participants denied any previous experience with exoskeletons. Physical condition was also asked for, and the majority of participants were sports students, so physical condition tended to be high.

For the M. deltoideus as well as for the M. trapezius significant reductions from ST to DT were observed (M. deltoideus: F (1, 8) = 8.672, p = 0.019, η_p² = 0.520; M. trapezius: ST–DT F (1, 8) = 17.268, p = 0.003, η_p² = 0.683). Moreover, for the exoskeleton use, the reduction of the muscle activity of the M. deltoideus failed to be significant (F (1, 8) = 4.849, p = 0.059, η_p² = 0.377) (cf. Figure 4). There were no differences between all examined conditions for the M. erector spinae (cf. Table 2 and Figure 4).

In addition to Table 2, Figure 4 shows the mean and standard deviation for each muscle in the percentage of MVC between pre- and post-fatigue and for single and dual tasks separately.

If also controlling for gender, post-hoc analysis showed significant differences for M. deltoideus across all conditions (F (1, 8) = 7.618, p = 0.025, η_p² = 0.488). Looking at the individual differences for the conditions, the results showed the following significant differences in the one-way ANOVA for the M. deltoideus: single-task post-fatigue without exoskeleton (F (1, 9) = 8.154, p = 0.021, η_p² = 0.505), single-task pre-fatigue with exoskeleton (F (1, 9) = 6.113, p = 0.039, η_p² = 0.433), dual-task pre-fatigue without exoskeleton (F (1, 9) = 6.295, p = 0.036, η_p² = 0.440), and dual-task post-fatigue without exoskeleton (F (1, 9) = 8.387, p = 0.020, η_p² = 0.512). For the conditions of single-task pre-fatigue without exoskeleton and dual-task pre-fatigue with exoskeleton, the results showed barely insignificant values.

Figure 3 shows the differences between the fatigue and the IDs (Fitts’ task) separately for the conditions with and without the exoskeleton. For the ST condition, no main effect of the exoskeleton was found (F (1, 8) = 0.112, p = 0.745, η_p² = 0.012), independent of the index of difficulty (ID) or the level of fatigue. A main effect of fatigue was found during ST conditions (F (1, 8) = 6.378, p = 0.032, η_p² = 0.415), regardless of ID or exoskeleton use (cf. Figure 4). In addition, a main effect was found for ST for the ID (F (1, 8) = 58.89, p < 0.001, η_p² = 0.867), with each ID leading to a significantly longer movement time than the lower one.

In the dual-task scenario, similar to single task, no main effect of exoskeleton was found (F (1, 8) = 0.837, p = 0.386, η _p² = 0.094), but a main effect of fatigue (F (1, 8) = 5.880, p = 0.041, η_p ² = 0.423) and ID (F (1, 8) = 30.116, p < 0.001, η _p² = 0.79). A main effect of task was found for the times to complete the Fitts’ tasks (F (1, 8) = 13.415, p = 0.005, η_p² = 0.598). The time was always higher in the dual-task conditions. Finally, neither a main effect nor an interaction (F (1, 8) = 1.793, p = 0.213, η_p² = 0.166) was found for the slopes of Fitts’ law calculated for each condition (pre and post-fatigue, with or without the use of an exoskeleton, during the single or dual task).

The number of errors between the four conditions (1) working without exoskeleton (mean errors 2.2 ± 1.4), (2) working with exoskeleton (mean errors 2.5 ± 2.3), (3) working without exoskeleton after fatigue (mean errors 2.3 ± 2.4) and (4) working with exoskeleton after fatigue (mean errors 0.8 ± 0.9) showed a mean difference (F (3, 7) = 4.618, p = 0.044, η_p² = 0.664). Post-hoc analysis showed that the number of errors in condition four was significantly lower than in all other conditions.

The results of the muscular activity showed diverging results, which are, however, consistent with the results of previous studies by Gillette et al. (33), which found a significant reduction in muscle activity of the M. deltoideus by using an upper-body exoskeleton, as well as with the review by Bär et al. (11). A 35% reduction in the activity of the M. deltoideus was observed, without considering the task or previous fatigue caused using the exoskeleton. In addition, a reduction of up to 52% in the activation of the M. trapezius and up to 23% in the activation of the M. erector spinae was also observed when using the exoskeleton, however, these observations failed to be significant. Therefore, we conclude that according to the results of the review by Bär et al. (11) that muscle activation might be task-specific and is also related to the weight of the nail gun tool.

Another potential reason might be the small sample size with an equal number of male and female participants. Within the study design of Kermavar et al. (34) especially the female participants had greater reductions in shoulder muscle activity. Regarding all task conditions, female participants showed an average of 12% higher muscle activation of the M. deltoideus. This could be due to the anthropometric differences between men and women, as exoskeletons today are still strongly adapted and designed for male anthropometry. Given the fact, that women have so far received little to no attention in research on supporting upper-body exoskeletons, this is an important step toward further understanding of the interaction between the body from an anthropometric view and an exoskeleton, which ideally adapts to the different anthropometric conditions and the needs of the user. Furthermore, it is necessary to understand these interactions to find out which measures need to be taken to optimally adjust the support of the female body, so that women can also benefit from exoskeletons.

A third explanation addresses the level of support. The review by Theurel and Desbrosses (20) showed that the use of a passive exoskeleton might induce unexpected consequences on the coordination between agonist and antagonist muscles. According to the findings by Kermavnar et al. (34), we were not able to determine a reduction in the activation of the M. erector spinae, but Walter et al. (35) found comparable results regarding the lack of significance between working with and without a passive exoskeleton with significant benefits of an active exoskeleton. This supports the idea that the level of support plays a decisive role, which needs to be examined in more detail in future studies in order to evaluate an optimal level of support (7).

Considering differences in muscular activity between single- and dual-task conditions, this study showed a significant reduction in the shoulder muscles of the M. deltoideus in the dual-task condition of 7–10% compared to the single-task condition, which could be an indication of lower motor control depending on the change in attentional resources (e.g., according to Wickens et al. (36)). Furthermore, reductions in DT conditions of up to 54% for the M. trapezius and up to 43% reduction in the M. erector spinae were observed using an exoskeleton even if there were no significant differences between the conditions (with and without exoskeleton) (cf. Table 2).

If we look at the differences between the use of an exoskeleton within the pre- and post-fatigue conditions, our study shows a reduction in muscular activation of the M. deltoideus in ST conditions of up to 26% without prior fatigue. With fatigue, there was an even greater effect of 32% reduction. Furthermore, a reduction of up to 53% in the muscular activation of the M. trapezius was shown when using the exoskeleton even without prior fatigue; with fatigue, the reduction was only 40%. For the M. erector spinae, the use of the exoskeleton also showed an effect without fatigue, with a reduction in muscular activation of up to 21% and with fatigue of up to 34%, which are similar to findings by Wei et al. (37) (cf. Table 2). This once again highlights the positively supportive effect of the upper-body exoskeleton on reduced muscular activation, particularly the enhanced supportive effect following a period of physical exertion, such as encountered in a workday characterized by overhead tasks in the industrial setting. This decrease in muscular activation signifies a diminished demand on the musculoskeletal system, especially the upper extremity and the shoulder girdle (21, 22). Consequently, using upper-body exoskeletons serves as a preventive factor for the reduction of work-related musculoskeletal disorders, especially in overhead tasks.

In summary, it is most reasonable to assume that the included participant sample size was too small to be able to determine a significant difference in the M. erector spinae for example. On the other hand, familiarization with the exoskeleton could also have played a decisive role, which, according to Walter et al. (35), is crucial for the positive impact of the exoskeleton. Our study population did not have any experience with upper-body exoskeletons prior to the beginning of this study and only underwent a brief familiarization phase at the start of the study. Nevertheless, this aspect reflects real-world scenarios, when workers will start using exoskeletons, as to this date most employees have no prior exoskeleton experience. This emphasizes the need for high-quality studies with trained and experienced exoskeleton users to demonstrate a significant reduction in muscular activity due to the potential benefits of the exoskeleton.

As predicted by the law of Fitts, drilling movement duration was linearly increased as the index of difficulty increased (24, 25). Indeed, the nature of the relationship between movement duration and task complexity is usually mathematically expressed as a linear relationship. However, this has been shown for table and pen tasks, such as linear pointing movements (25) or circular drawing tasks (29). In general, it has been known for more than two decades that a wide variety of rapid aimed movements does respect Fitts’ law (38). However, the use of ecological tasks is very rare in the literature, and Fitts’ law has never been evaluated in drilling tasks such as in the present study. Therefore, it is interesting to note that Fitts’ tasks usually performed in a laboratory setting on a paper-to-pen test can be transferred to an ecological upper-body task. This opens a wide new range of research in the field of ergonomics and the study of motor control of work-related tasks. This is utterly important since daily activities involve fine motor skills to a greater extent than standardized laboratory tasks and involve a higher level of cognitive process (39).

Interestingly, the present study also showed the feasibility of using a Fitts’ task while wearing an exoskeleton, since the Fitts’ law was still respected. The observation is even more striking, since the use of an exoskeleton did not affect the Fitts’ relationships, whatever the level of fatigue or cognitive requirement. This was expressed by identical movement durations in all indexes of difficulties and slopes of Fitts’ relationships. In patients, it is known that wearing an exoskeleton may alter movement accuracy, especially with the heaviest exoskeletons (40). Conserving a smooth movement using an exoskeleton, whether it is a passive or active model, often represents a huge challenge for companies in all fields, including industrial and medical (41). The present study raised that (1) healthy young participants can easily adapt to the use of an exoskeleton to perform a fast and accurate task, and (2) wearing an exoskeleton (at least for the model used in the present study), does not alter the ability to manage the speed-accuracy trade-off.

In the present study, Fitts’ law was also not affected by muscle fatigue in its form, but a global decrease in MT has been observed. It has been shown that muscle fatigue was more prone to induce a global increase in movement time during a classical Fitts’ pointing task (26). However, in the study of Missenard et al., a decrease in muscle force of 30% was necessary to induce a change in movement duration. Therefore, it is possible that the present study did not induce sufficient muscle fatigue to decrease muscle performance. Moreover, it is known that there is a specificity of the fatiguing exercise to the following changes observed in the neuromuscular system (42). In other terms, the lack of direct correspondence of the fatiguing exercise and the post-task (Fitts’ drilling task), despite a correspondence in solicited muscle groups, may lead to a lack of muscle fatigue in the present study, or even the opposite, an increase in performance (e.g., a decrease in movement time in Fitts’ task). Then, this latter may be due to an overcompensation or a change in muscle activation strategies. It is well known that the motor system can adapt its strategy to compensate for muscle fatigue, to preserve the Fitts’ law behavior, as observed by a similar slope before and after fatigue (26). In other words, after a fatiguing exercise, the Fitts’ law could be altered only quantitatively, but not qualitatively. Interestingly, Missenard et al. did not find any changes in EMG activity of agonist or antagonist muscles between pre- and post-fatigue, despite a global increase in movement duration during the Fitts task. Again, this shows the ability of the system to adapt to any fatiguing exercise, and that neither EMG nor movement parameters could solely be considered to account for neuromuscular fatigue.

Finally, a globally greater movement duration was observed in dual task as compared to single task, as evidenced by an upward shift of Fitts’ law in all conditions (with/without exoskeleton, with/without fatigue). This shows that the lack of cognitive resources may lead to an increase in movement time, whatever the index of difficulty of the task. Resource allocation to the DT may lead the motor system to make the choice to slow down the movement to be able to process the cognitive task properly. It has been shown for instance that in the presence of mental fatigue, a state that also generates a decrease in cognitive resources, an upward shift in Fitts’ law (e.g., a global increase in movement duration) was also observed (27). According to motor control theories, the central nervous system, mainly at the stage of motor planning, may adopt this strategy of slowing down the movement to preserve task success (43). However, it should be noted that, despite this upward shift, Fitts’ slope relationship was not altered by the dual task. Then, after muscle fatigue, the shape of the Fitts’ law was altered quantitatively but not qualitatively. Therefore, this shows again that the system could adjust movement duration similarly for each index of difficulty to preserve its ability to manage the speed-accuracy trade-off. For generalization, these results should be confirmed with larger sample sizes.

In contrast to the results of Govaerts et al. (44), Bequette et al. (13), Federici et al. (14), and Zhu et al. (15), our results of cognitive performance, measured by the error frequencies in the different conditions, clearly show a clear advantage in working with an exoskeleton, especially after physical fatigue, which should be comparable to a working day following the fatigue protocol by Bär et al. (11), and Leone et al. (12). A 65% reduction in errors was recorded compared to working without an exoskeleton (see the section “Cognitive performance (errors) within the test conditions”). This could be a compensatory effect of the exoskeleton, or the complaints were not perceived in the fatigue state. One might now also assume that learning effects could possibly be responsible for this reduction, but this can be rejected as the trials were carried out in randomized order. In addition, no task prioritization effect could be observed based on the results of the Fitts’ task.

In summary understand all changes in cognitive-motor control and resource allocation discussed with respect to our study findings we recommend using study designs, such as those by Zhu et al. (15), to examine brain activity (e.g., fNIRS, EEG) to gain more insight into the shift in the attentional focus.

In addition, further studies based on the muscle synergy approach are much needed, as this method considers how different muscle groups interact to produce coordinated movements. Following this framework will allow a better understanding of the impact of wearing an exoskeleton in terms of movement coordination, the control of specific muscle groups, and the cognitive processes involved in managing these movements. Since these factors have a significant influence on the effectiveness of exoskeleton use, further research will provide valuable insights for the optimization of exoskeleton designs and their use in different applications. Finally, it would be beneficial to also monitor heart rate or heart rate variability within the experimental design to get insights into physical exhaustion that might impact the different performance levels.

The findings of this study furthermore underline the importance of a holistic view of exoskeleton usage in the field in which not only its potential physiological benefits during motion activities are to be considered, but also how this human–machine coupling will affect other areas such as attentional resources and thereby productivity in the long run as well. Exoskeleton integration into the workplace should therefore not only aim for a general familiarization of its usage but furthermore identify areas in which exoskeletons can have the potential of lowering attentional stress (or place an additional burden for that matter) in order to maintain productivity in conjunction with workplace safety.