Sage Journals: Discover world-class research

Abstract

Objective

To investigate how varying workload intensity and personalized conditions influence physiological stress responses and task efficiency in human–robot collaboration (HRC).

Background

HRC is increasingly used to enhance productivity and reduce physical demands, yet workers’ mental and physiological strain remains unaddressed. High workload intensity, often induced by robot pacing, can elevate autonomic stress responses, and current systems rarely adapt workload to individual worker capacity. Understanding how workload levels, particularly personalized pacing, affect physiological stress and task performance is essential for designing human-centric collaborative workplaces.

Methods

The study consisted of two experiments. In the first experiment (E1), participants completed an assembly task with a collaborative robot (CR) under two workload conditions (low, high). In the second experiment (E2), participants completed an additional personalized scenario in which the CR motion parameters were adjusted to each participant’s capacity. Heart rate variability (HRV) indicators were assessed, and task performance was evaluated using assembly time (AT).

Results

In E1, the high workload reduced AT, but increased the heart rate (HR) and decreased the root mean square of successive differences (RMSSD), indicating elevated stress. In E2, the personalized scenario maintained efficiency while lowering the HR and restoring the RMSSD compared to the high-intensity condition. The high-frequency (HF) power showed no significant variation in either experiment.

Conclusions

Workload intensity significantly affects both efficiency and physiological stress in HRC. Personalizing the CR motion parameters to worker capacity preserves performance while reducing stress responses.

Application

Adaptive, capacity-based HRC strategies can help manufacturers sustain productivity without compromising worker well-being.

Keywords

human-centric design health physiological stress heart rate variability industry 5.0

Introduction

Human–robot collaboration (HRC) has become an increasingly important operational model in modern manufacturing, where human flexibility and decision making are combined with the consistency and repeatability of robotic systems. Numerous studies have shown that well-designed collaborative systems can significantly improve productivity and reduce physical ergonomic risks (Montini et al., 2024; Villani et al., 2018). Despite these advancements, the human component of collaboration remains highly sensitive to factors such as task pacing, cognitive load, and physiological stress, elements that are often insufficiently addressed in current industrial applications (Hopko et al., 2022; Panagou et al., 2024). Recent discussions on human-centric manufacturing have emphasized the importance of aligning advanced technologies with worker needs and well-being. Although such ideas are frequently associated with the Industry 5.0 framework (Breque et al., 2021), many of its principles, including collaborative intelligence, ethical automation, and sustainable workplaces, represent an evolution of concepts already present within Industry 4.0. Rather than interpreting Industry 5.0 as a distinct technological revolution, the present work adopts it as a broader contextual perspective that reflects an ongoing movement toward more human-aligned system design. This framing avoids overstating the novelty of the Industry 5.0 narrative while acknowledging the relevance of human-centric values in emerging industrial systems. One of the main challenges in HRC is the insufficient integration of psychological and physiological well-being indicators into manufacturing system assessment frameworks.

Many systems still rely heavily on physical ergonomic models, overlooking crucial aspects of mental workload, stress response, and cognitive fatigue (Simões et al., 2022; Su et al., 2024). This omission is significant, as cognitive strain directly affects productivity, safety, and long-term health (Halkos & Bousinakis, 2010). According to the World Health Organization (2022), more than 12 billion working days are lost annually due to stress-related conditions, resulting in over $1 trillion in global economic losses. To address these challenges, researchers have increasingly combined subjective and objective methods for evaluating workload in HRC. Subjective tools such as the NASA Task Load Index (NASA-TLX) are frequently used to capture perceived mental workload (Hart & Staveland, 1988). However, subjective assessments remain vulnerable to interpersonal bias and retrospective inaccuracies (Cohen, 2013). Objective physiological indicators, particularly heart rate variability (HRV), offer a complementary perspective by capturing autonomic nervous system responses to cognitive and emotional demands (Shaffer & Ginsberg, 2017). A comprehensive review by Sharma and Gedeon (2012) highlights the range of available stress-measurement modalities, including invasive biochemical markers, noninvasive autonomic measures such as electrocardiography (ECG), and central nervous system indicators such as electroencephalography (EEG). Among HRV parameters, heart rate (HR), the root mean square of successive differences (RMSSD), and high-frequency (HF) power are widely recognized for their sensitivity to psychological stress and physical strain (Baevsky & Berseneva, 2008; Borghi et al., 2024). Increased HR and decreased RMSSD typically indicate elevated sympathetic activity, while HF reflects parasympathetic activation and the capacity for physiological recovery (Shaffer & Ginsberg, 2017). Several recent studies have attempted to align workload research with practical industrial constraints. Coronado et al. (2022) classified performance and well-being metrics for HRC, highlighting the lack of comprehensive real-world validation. Su et al. (2024) examined the impact of robot speed on worker mental state but did not incorporate individual variability in execution capacity. To overcome such limitations, emerging research has proposed adaptive task allocation and personalized robot behaviors (Bouaziz et al., 2024; Calzavara et al., 2024). These approaches aim to avoid underload, associated with monotony and disengagement, as well as overload, which leads to fatigue and physiological strain. However, empirical evidence that jointly evaluates performance outcomes and physiological stress response across varying workload intensities remains scarce, particularly in realistic HRC settings (Panchetti et al., 2023; van Dijk et al., 2023).

The present study addresses these gaps by introducing a multiscenario experimental framework that manipulates collaborative robot (CR) motion parameters to create distinct workload-intensity conditions in a collaborative assembly task. In this context, workload intensity refers to the proportion of time in which a worker is actively engaged in task execution, while participant capacity denotes the individual’s baseline execution pace measured in a preliminary skill test. Building on prior subjective workload findings obtained using NASA-TLX in our previous work (see Javernik et al., 2023), and recent work on adaptive, human-centric system design (Ojsteršek et al., 2024), the present study focuses on objective ECG-based HRV measures to investigate how robot-induced pacing influences physiological stress responses and task performance. NASA-TLX was therefore not re-administered or re-analyzed in this study, as the subjective workload patterns have already been reported comprehensively in Javernik et al. (2023).

By evaluating low-intensity, high-intensity, and adjusted-intensity (personalized) scenario, the study aims to clarify the trade-offs between performance and physiological stress response, and to assess whether adjusting robot pacing to individual capacity can preserve efficiency while reducing stress. Overall, this research contributes to a more comprehensive understanding of human performance and well-being in HRC, offering empirical evidence to support the development of more adaptive, human-aligned collaborative systems.

Purpose of Research

The purpose of this research was to examine how different levels of workload intensity influence workers’ physiological stress responses and task efficiency during HRC. The study evaluates objective indicators of worker well-being, with a primary focus on cardiac measures, including HR and HRV parameters, specifically RMSSD, and HF. Task efficiency was assessed through assembly times (ATs) recorded during each collaborative scenario. To investigate these effects, the study incorporated multiple workload-intensity conditions by manipulating the motion parameters of a CR. Importantly, experimental design evolved during the study: an initial version (Experiment 1; E1) included low-intensity and high-intensity scenarios, while subsequent findings revealed that neither condition fully reflected a human-centric collaborative setting. This led to the development of a modified design (Experiment 2; E2), which introduced an additional scenario tailored to each participant’s individual task execution capacity. The research therefore pursued two complementary aims:

(1) to determine how low versus high workload intensity affects physiological stress response and performance in a controlled HRC task (E1), and

(2) to evaluate whether personalizing workload intensity to individual capacity can reduce stress while maintaining efficiency (E2).

Each experiment addressed these aims separately, and hypotheses were formulated to correspond to the specific structure of E1 and E2.

Methods

Independent Variables

In this study, workload intensity was the independent variable. It was defined as the level of engagement of the human operator during collaboration and it was directly linked to the utilization level of each participant. Workload intensity was determined by the motion parameters of the CR, as its speed and acceleration settings governed the temporal dynamics of the collaborative operation and the proportion of time that participants were actively involved in the work cycle.

During the experiment, participants were exposed to three levels of workload intensity: low, high, and personalized. At the low-intensity level, the CR motion parameters were adjusted to provide a slower operational rhythm, allowing participants to be actively engaged for approximately 60% of the total cycle time. At the high-intensity level, the CR operated at a substantially faster pace, resulting in continuous engagement of the participant (full utilization) and a state of task overload. Under these conditions, maintaining synchronization with the CR would have required participants to exceed their capacity by approximately 120%. Finally, in the personalized condition, the CR motion parameters were tuned to match each participant’s individual capacity, ensuring full utilization without exceeding personal limits. In practical terms, this meant that the robot and the human reached the shared assembly point at approximately the same time, eliminating both unnecessary idle waiting and the overload experienced in the high-intensity condition. This synchronization created a balanced collaborative flow in which neither partner fell behind nor was pushed beyond their execution capability.

Because individuals differ in their capacity, a preliminary test was conducted to ensure comparable experimental conditions across participants. Prior to the experimental session, each participant completed a 90-s manual skill test, which replicated the collaborative assembly operation. The result of this test was the average normative time per product, which served as the basis for defining the CR motion parameters for each participant at all three workload levels. Given the simplicity of the assembly operation, the relatively short duration of the preliminary test was sufficient to capture a reliable estimate of each participant’s individual capacity and to ensure that the CR motion parameters were precisely adjusted to achieve the intended workload intensity levels.

Dependent Variables

Two categories of dependent variables were examined in this study: physiological stress response indicators and task performance measures. Physiological responses were assessed by HR and HRV parameters, while performance was evaluated based on the average AT.

HRV serves as a key indicator of the balance between the two autonomic nervous systems (parasympathetic and sympathetic) and provides valuable physiological measures for assessing human health and the stress response (Shaffer & Ginsberg, 2017). The assessment of HRV begins with the acquisition of an ECG, a noninvasive method that records the electrical activity of the heart via electrodes placed on the skin. By analyzing the recorded ECG signal, the precise timing of each heartbeat (R-peaks) can be determined. The intervals between successive R-peaks, known as RR intervals, form the basis for HRV analysis. By monitoring HRV, researchers can gain insights into a worker’s autonomic balance, stress response, and overall physiological adaptation, all of which influence both well-being and productivity.

In this study, the RR intervals were recorded using a Bittium Faros 180 ECG device and analyzed with Kubios HRV software, a widely used tool for HRV signal processing (Tarvainen et al., 2014). Kubios’ automatic artifact detection was applied, and recordings containing more than 1% artifacts were excluded. All accepted recordings were visually inspected to confirm correct R-peak identification. For each workload condition, one continuous 10-min artifact-free epoch was extracted. To avoid potential distortion of HRV parameters caused by age-related cardiac irregularities, only younger adults were included in the sample. This approach enabled the reliable extraction of HRV parameters reflecting the participants’ autonomic balance, stress response, and overall physiological adaptation during HRC.

According to Shaffer and Ginsberg (2017), various parameters are used to assess HRV in experimental settings, each capturing distinct aspects of autonomic function. These include time-domain measures, frequency-domain measures, and nonlinear metrics. We focused on the analysis of three specific HRV parameters, as they offer meaningful insights into stress regulation and autonomic nervous system:

HR, as a fundamental physiological parameter, represents the number of heartbeats per minute. An elevated HR is typically associated with increased sympathetic activity, signaling physiological arousal in response to stress or task demands. While HR alone does not provide detailed information on autonomic balance, it serves as a complementary measure to the HRV indices.

RMSSD is a time-domain measure that quantifies short-term variations in the RR intervals. It is particularly sensitive to parasympathetic activity, reflecting vagal tone and the ability to recover from stress. Higher RMSSD values indicate greater autonomic flexibility and resilience to stress.

HF power is a frequency-domain measure that captures heart rate oscillations within a specific frequency range (typically 0.15–0.40 Hz). It is influenced predominantly by parasympathetic activity and is linked closely to respiratory sinus arrhythmia, the natural fluctuation of the heart rate with breathing. Increased HF power suggests enhanced parasympathetic modulation and a greater capacity for relaxation and stress mitigation.

Due to the complexity of the HRV parameters, a simplified overview of their states and relation to physiological stress is shown (Table 1).

Table 1.

HRV Parameters and Their Relationship to Physiological Stress.

Parameter	Increase	Decrease
Heart rate (HR)	Higher stress	Lower stress
High-frequency (HF) power	Lower stress	Higher stress
Root mean square of successive differences (RMSSD)	Lower stress	Higher stress

In addition to the physiological measures, AT was analyzed as a key indicator of task performance and operational efficiency during collaboration with the CR. AT represents the time required by a participant to complete one full assembly cycle, including all manual subassembly actions and the interaction phase with the CR. For each workload scenario, an average AT was calculated for every participant, based on the assembly durations recorded across all completed work cycles within that scenario.

To obtain reliable and precise measurements, ATs were extracted from video recordings of the collaborative operation. Each participant’s work session was filmed, and the start and end points of each assembly cycle were identified manually during video analysis. Shorter average ATs indicate higher task efficiency and better synchronization with the CR, whereas longer times reflect reduced performance, hesitation, or challenges in maintaining coordination at a given workload intensity.

Experiment 1

Participants

A total of 44 volunteers participated in the first experiment (E1), including 33 men and 11 women, with a mean age of 25.5 years (SD = 2.92). One participant was excluded due to data inconsistencies in the physiological recordings, resulting in a final sample of 43 participants. The participants were healthy younger adults with no self-reported medical conditions, recruited from the local university community and consisting primarily of students and early career researchers with no prior experience with the CR or the specific assembly task used in the study. All participants were informed about the study procedures and signed a written informed consent form prior to participation. Participation was voluntary and uncompensated.

Human–Robot Collaboration Setup and Assembly Operation

The experiment was conducted in a controlled laboratory environment configured as a collaborative workplace. The collaborative workplace consisted of a CR UR3e ① equipped with a Robotiq gripper 2F-85 ②, a worktable ③, a switch with light indicators ④ and button ⑤, a preparation area ⑥, a buffer of semi-finished products ⑦, ⑧, ⑨, and finished products ⑩ (Figure 1).

Figure 1.

Layout of the collaborative workplace used in the experiment.

The collaborative assembly task required participants and the CR to jointly assemble a product composed of three semi-finished products (SFPs). SFP1 was represented by a standard LEGO Duplo brick, whereas SFP2 and SFP3 were custom-designed components that increased the complexity of the manual assembly portion of the task (Figure 2). The assembly operation followed a cyclic sequence. First, the CR ① picked up SFP1 from its designated buffer and transported it to the shared assembly point (shared workspace location). In parallel, the participant worked on manually preparing two SFP4, each consisting of one SFP2 and one SFP3. When the CR reached the assembly point with SFP1, the light indicator on switch ④ turned green, indicating that the participant could attach the two prepared SFP4 to SFP1. After completing the attachment, the participant pressed the confirmation button ⑤, switching the indicator to red ④ and signaling the CR to proceed. The CR then transported the fully assembled product to the buffer of finished product ⑩, while the participant immediately began preparing the next pair of SFP4s.

Figure 2.

Components used in collaborative assembly operation.

Experimental Design

E1 was designed to study how different levels of workload intensity during HRC affect participants’ physiological stress responses and task efficiency. A within-subject design was applied, each participant completed two collaborative scenarios differing in workload intensity: a low-intensity scenario (S1) and a high-intensity scenario (S2) (Figure 3).

Figure 3.

Design of E1.

In the introduction phase of the experiment, prior to collaboration with the CR, participants were familiarized with the experimental sequence and collaborative operation, had the ECG device attached to monitor their HRV throughout the experiment, and completed a 90-s manual skill test. As described in the Independent Variables section, the skill test was used to determine the CR motion parameters for each participant’s individual capacity, ensuring that workload intensity levels were comparable across participants.

After the introduction phase, participants completed the two HRC scenarios in a fixed order, with a five-minute break in between. In S1, the CR operated at a slower pace, resulting in a low workload intensity, where participants were actively engaged for approximately 60% of the work cycle. In S2, the CR motion parameters were substantially increased to create a high-intensity condition. Under these parameters, participants were engaged continuously, and maintaining full synchronization with the CR would have required them to exceed their individual capacity by approximately 20%, creating a state of task overload. Physiological data (HRV parameters) and task performance data (AT) were recorded in each scenario to enable direct within-subject comparison between low and high workload intensity.

Statistical Methods

To evaluate the effects of the two workload intensity levels in E1, a within-subject statistical approach was applied. Because the distributions of some dependent variables (RMSSD and AT) violated the normality assumption according to the Shapiro–Wilk test, and given the relatively small sample size, a nonparametric Wilcoxon signed-rank test was used as an alternative to the paired t-test. The test compared paired observations from the low-intensity scenario (S1) and the high-intensity scenario (S2).

All tests were two-tailed, and the level of statistical significance was set at α = .05. Effect sizes for the Wilcoxon tests were reported as r, derived from the standardized test statistic. The hypotheses (Table 2) were formulated such that the alternative hypotheses predicted that the higher workload intensity in S2 would result in increased physiological stress response (higher HR, lower RMSSD, lower HF) and improved efficiency (shorter AT) relative to S1.

Table 2.

Null and Alternative Hypotheses for the Wilcoxon Signed-Rank Test on Physiological and Performance Variables.

HR	H₀	The median of the differences between the HR_S2 and HR_S1 is lower than zero.
HR	H₁	The median of the differences between the HR_S2 and HR_S1 is greater than zero.
RMSSD	H₀	The median of the differences between the RMSSD_S2 and RMSSD_S1 is greater than zero.
RMSSD	H₁	The median of the differences between the RMSSD_S2 and RMSSD_S1 is lower than zero.
HF	H₀	The median of the differences between the HF_S2 and HF_S1 is greater than zero.
HF	H₁	The median of the differences between the HF_S2 and HF_S1 is lower than zero.
AT	H₀	The median of the differences between the AT_S2 and AT_S1 is greater than zero.
AT	H₁	The median of the differences between the AT_S2 and AT_S1 is lower than zero.

Results

The descriptive statistics (Table 3) show that the mean HR increased from 78.49 bpm in S1 to 82.10 bpm in S2. In contrast, the mean value of RMSSD decreased from 33.71 ms in S1 to 27.01 ms in S2. HF power remained practically unchanged, with values of 21.60 ms² in S1 and 21.68 ms² in S2. Performance, measured through mean AT, improved under the higher workload intensity, as the mean AT decreased from 11.91 s in S1 to 9.46 s in S2.

Table 3.

Descriptive Statistics for E1.

Parameter	Unit	M	SD	Md
HR (S1)	(bpm)	78.49	13.19	79.87
HR (S2)	(bpm)	82.10	12.82	83.95
HF (S1)	(ms²)	21.60	9.61	20.46
HF (S2)	(ms²)	21.68	10.98	21.44
RMSSD (S1)	(ms)	33.71	17.73	28.75
RMSSD (S2)	(ms)	27.01	14.42	22.79
AT (S1)	(s)	11.91	1.58	11.79
AT (S2)	(s)	9.46	1.10	9.44

To determine the statistical significance of these differences, a Wilcoxon signed-rank test was conducted (Table 4). HR was significantly higher in S2 (Z = 5.156, p < .001, r = .87). No significant difference was found in HF (Z = −0.193, p = .847, r = .02). RMSSD was significantly lower in S2 (Z = −5.712, p < .001, r = .82), and ATs were significantly shorter (Z = −5.385, p < .001, r = .82).

Table 4.

Related-Samples Wilcoxon Signed-Rank Test Summary.

Pairwise Comparisons	Sample Size (N)	W	z	p	r
HR_S2–HR_S1	43	900	5.156	<.001	.87
HF_S2–HF_S1	43	457	−0.193	.847	.82
RMSSD_S2–RMSSD_S1	43	0	−5.712	<.001	.02
AT_S2–AT_S1	43	27	−5.385	<.001	.82

For clarity, the Wilcoxon signed-rank test results are presented graphically (Figure 4). The histograms illustrate the number of participants exhibiting positive (green) and negative (red) differences for each HRV indicator. Thirty-eight out of the 43 participants showed an increased mean HR in S2 (Figure 4(a)). The distribution of HF differences is almost balanced, with 21 participants showing positive differences and 22 showing negative differences (Figure 4(b)). Regarding RMSSD, 39 of the 43 participants exhibited a decrease in S2 (Figure 4(c)). In terms of efficiency, all the participants achieved a shorter mean AT in S2 compared to S1 (Figure 4(d)).

Figure 4.

Distribution of the differences between S2 and S1 (Wilcoxon signed-rank test) in: (a) HR, (b) HF, (c) RMSSD, and (d) AT.

Discussion

The results of E1 show that increased workload intensity affected participants’ physiological state and their task performance. HR increased significantly in S2, with 38 of the 43 participants showing higher mean HR, indicating elevated physiological activation. Similarly, RMSSD values decreased significantly from S1 to S2, with 39 participants exhibiting lower values in S2. A reduced RMSSD is generally associated with decreased parasympathetic activity and higher physiological strain and stress response, suggesting that participants experienced a greater stress response during high-intensity collaboration. Together, the combination of higher HR and lower RMSSD thus provides convergent evidence that S2 imposed substantially greater stress demands. The large effect sizes observed for HR and RMSSD reinforce the magnitude of this physiological shift, confirming that the high-intensity condition had a strong autonomic impact.

In contrast, HF power did not differ significantly between scenarios. The nearly balanced distribution of positive and negative HF differences (21 vs. 22 participants) suggests that HF may be less sensitive to short-term fluctuations in robot-induced pacing or that the exposure duration in E1 was insufficient to elicit detectable changes in respiratory-linked parasympathetic modulation.

Performance outcomes showed the opposite pattern: all participants achieved faster ATs in S2 compared to S1. The very large effect size for AT further demonstrates that performance gains were robust under high workload intensity. The decrease in AT reflects increased operational efficiency under the higher CR pace, although this improvement occurred alongside elevated physiological stress response. Thus, E1 reveals a clear trade-off in HRC: higher workload intensity enhances task efficiency but concurrently increases physiological load.

Overall, the findings from E1 support all corresponding alternative hypotheses (H₁) except for HF, for which the null hypothesis was retained. The consistent changes in HR, RMSSD, and AT indicate that manipulating CR motion parameters to increase workload intensity strongly influences both participants’ well-being and performance. These results highlight the importance of considering individual physiological responses when designing collaborative work systems, as elevated intensity can quickly improve performance while increasing stress response. Such insights emphasize the need for more balanced or adaptive approaches in HRC and validate the experimental design of E1 as an effective foundation for further refinement. This limitation motivated the development of the second experiment (E2), where an additional personalized scenario was introduced to examine whether individually adjusted workload intensity may better balance efficiency and physiological responses.

Experiment 2

Participants

The nineteen participants in E2 consisted of 11 men and 8 women, with a mean age of 26.5 years (SD = 3.10). They represented the final subset of the original 44 participants from E1 and completed the additional scenario (S3) immediately after finishing the E1 procedure. Because they had already completed all required familiarization steps in E1, no additional training was needed for E2. The smaller sample size in E2 resulted from the need to refine and extend the experimental design that emerged during the ongoing data collection in E1. The eligibility characteristics were identical to those described in the Participants subsection of E1, and the same ethical approval and consent procedures applied.

Experimental Design

E2 followed the same overall experimental procedure as described in E1, except that an additional scenario, Scenario 3 (S3), was added (Figure 5). A human-centric approach was applied in S3, with the CR motion parameters adjusted to align with each participant’s individual capacity. As outlined in the Independent Variables section, S3 served as an adjusted workload condition intended to explore whether such a participant-tailored collaboration tempo could lead to a better balance between worker well-being and task efficiency, avoiding the underload observed in S1 and the overload present in S2.

Figure 5.

Design of E2.

Statistical Methods

For E2, the Friedman test was used to evaluate the effects of three workload intensity scenarios on physiological and performance measures. Again, a nonparametric test was selected because the assumption of normality for the RMSSD, HF, and AT variables was not met, as indicated by Shapiro–Wilk test of normality, and because the sample size in E2 was even smaller than in E1.

The Friedman test was conducted separately for each dependent variable to determine whether any overall differences existed across the three scenarios. When a significant overall effect was detected, pairwise comparisons were performed using Wilcoxon signed-rank tests (S1 vs. S2, S1 vs. S3, and S2 vs. S3) to identify where the differences occurred. Adjustments for multiple comparisons were applied, and all tests were two-tailed with a significance level of α = .05. In accordance with the structure and assumptions of the Friedman test, hypotheses were designed (Table 5). To complement the significance testing in E2, effect sizes were also computed. For the Friedman test, effect size was expressed using Kendall’s W (W) and interpreted according to conventional thresholds (small = .10, moderate = .30, large = .50; Cohen, 1992). For the post-hoc Wilcoxon signed-rank comparisons, effect sizes were calculated using the standardized test statistic (r).

Table 5.

Null and Alternative Hypotheses for the Friedman Test on Physiological and Performance Variables.

HR	H₀	There is no significant difference in the HR across the three scenarios.
HR	H₁	At least one scenario differs in the HR from the other scenarios.
RMSSD	H₀	There is no significant difference in the RMSSD across the three scenarios.
RMSSD	H₁	At least one scenario differs in the RMSSD from the other scenarios.
HF	H₀	There is no significant difference in the HF across the three scenarios.
HF	H₁	At least one scenario differs in the HF from the other scenarios.
AT	H₀	There is no significant difference in the AT across the three scenarios.
AT	H₁	At least one scenario differs in the AT from the other scenarios.

Results

The descriptive statistics (Table 6) show that the mean HR increased from 77.69 in S1 to 82.10 bpm in S2, and then decreased in S3 to 79.41 bpm, partially returning toward the level observed under low-intensity conditions. The HF remained relatively stable across the scenarios, with only minimal fluctuations. The difference between S1 and S2, and S2 and S3, was approximately 3%, while almost no difference appeared between S1 and S3. The RMSSD followed a similar trend as HR in terms of physiological interpretation, declining from 34.59 ms in S1 to 27.36 ms in S2, then increasing slightly to 29.93 ms in S3. This pattern suggests an elevated physiological stress response in S2 compared to S1, and partial recovery in S3 compared to S2. Regarding the participants’ performance, the mean AT decreased from 11.32 s in S1 to 9.21 s in S2, and remained similarly low in S3 with 9.31 s, indicating that task efficiency was maintained under the personalized condition. From a descriptive perspective, these results suggest that high efficiency can still be achieved under conditions adjusted to the participants’ capabilities, while simultaneously reducing the physiological stress response.

Table 6.

Descriptive Statistics for E2.

Parameter	Unit	M	SD	Md
HR (S1)	(bpm)	77.69	14.10	75.05
HR (S2)	(bpm)	82.10	13.39	83.95
HR (S3)	(bpm)	79.41	13.93	76.55
HF (S1)	(ms²)	20.57	8.40	20.46
HF (S2)	(ms²)	21.19	9.72	20.21
HF (S3)	(ms²)	20.55	10.60	18.74
RMSSD (S1)	(ms)	34.59	17.19	34.52
RMSSD (S2)	(ms)	27.36	13.90	24.34
RMSSD (S3)	(ms)	29.93	16.56	27.61
AT (S1)	(s)	11.32	1.17	11.62
AT (S2)	(s)	9.21	0.96	9.44
AT (S3)	(s)	9.31	1.01	9.55

A related-samples Friedman test was conducted to determine whether the observed differences in physiological and performance parameters across the three scenarios were statistically significant. The results (Table 7) revealed statistically significant differences in HR (χ² (2) = 18.11, p < .001, W = .48), RMSSD (χ² (2) = 18.42, p < .001, W = .48), and AT (χ² (2) = 29.79, p < .001, W = .78). In contrast, the HF showed no significant difference across scenarios (χ² (2) = 0.11, p = .949, W = .00), supporting the earlier observations that the HF remained relatively unaffected by the changes in workload intensity. These findings are consistent with the descriptive results, where HR and RMSSD reflected clear changes in physiological state as workload intensity varied, and AT showed substantial improvements in task efficiency under higher or personalized conditions, while HF remained stable across all three scenarios.

Table 7.

Related-Samples Friedman’s Two-Way Analysis of Variance by Ranks.

Parameter	Sample Size (N)	Test Statistic (χ²)	Degree of Freedom (df)	Asymptotic Sig. (p)	W
HR	19	18.105	2	<.001	.48
HF	19	0.105	2	.949	.00
RMSSD	19	18.421	2	<.001	.48
AT	19	29.789	2	<.001	.78

To understand the significant differences identified by the Friedman test more clearly, the distribution of mean ranks for each parameter across the three scenarios is presented visually (Figure 6). The Friedman test ranks each participant’s values across the scenarios and then analyzes these ranks rather than the raw scores (Friedman, 1937). Since the experiment included three scenarios, each value was assigned a rank from 1 to 3, where higher values received higher ranks. Accordingly, scenarios with higher mean ranks reflect higher overall values of the corresponding parameter.

Figure 6.

Distribution of ranks across the scenarios for (a) HR, (b) HF, (c) RMSSD, and (d) AT based on the related-sampled Friedman’s test.

The rank distributions and corresponding mean ranks offer further insight into the differences between the physiological and performance measures across the three scenarios. For HR, the highest mean rank occurred in S2, indicating that participants experienced the strongest physiological activation under the high-intensity condition (Figure 6(a)). In S3, the mean HR rank was lower than in S2 and closer to the value observed in S1, which is consistent with the partial reduction in HR seen in the descriptive statistics. The mean ranks for HF were almost identical across all scenarios (Figure 6(b)), reinforcing the nonsignificant differences reported earlier. The rank distribution of RMSSD followed a similar trend to HR in terms of physiological stress response (Figure 6(c)). The higher workload intensity corresponded to greater physiological stress response, as most participants achieved the highest rank in S1 and the lowest rank in S2. As with HR, the RMSSD in S3 shows a balance between S1 and S2, representing the extremes of workload intensity. For task performance, the mean ranks for AT clearly reflected the influence of workload intensity. The lowest mean rank was observed in S2, demonstrating that most participants achieved their shortest ATs under the high-intensity condition. The mean rank in S3 was similar to that of S2, indicating that efficiency was maintained under the personalized condition. In S1, the mean rank value was 3.00, meaning that all the participants in this scenario recorded the longest ATs across the three scenarios (Figure 6(d)).

Post-hoc pairwise comparisons with the Bonferroni correction were conducted to further examine the significant differences between HRC scenarios, as identified by the Friedman test for HR, RMSSD, and AT. These comparisons provide a detailed understanding of how each HRC scenario influenced the workers’ stress response and efficiency.

For HR, the results revealed significant differences between S1 and S2 (adj. P < .001, r = .97) and between S2 and S3 (adj. P = .028, r = .60; see Table 8). These findings confirm that the high-intensity scenario elicited a greater physiological stress response than both the low-intensity and personalized conditions. Although participants were utilized 100% in both S2 and S3, the adjustment of workload intensity to individual capacity in S3 resulted in a significantly lower HR compared to S2. No significant difference was found between S1 and S3 (adj. P = .314, r = .37), indicating that the personalized conditions in S3 did not result in significantly elevated HR relative to the low-intensity scenario (S1).

Table 8.

Pairwise Comparisons of HR Across the Scenarios Using Post-Hoc Analysis From the Friedman Test.

Pairwise Comparisons	Std. Test Statistic	Sig.	Adj. Sig. (p)	r
HR_S1-HR_S3	−1.622	.105	.314	.37
HR_S1-HR_S2	−4.218	<.001	.000	.97
HR_S3-HR_S2	2.596	.009	.028	.60

Regarding HF, none of the pairwise comparisons showed significant differences between the scenarios (all adj. P = 1.000; see Table 9). These results reinforce conclusions from both the Friedman test and the descriptive statistics, suggesting that HF is less sensitive to, or is not influenced by, workload intensity variations in the context of HRC, or that the experimental sample size was too small to reveal such effects. Consistent with this, all HF pairwise comparisons showed negligible effect sizes (r values below .10).

Table 9.

Pairwise Comparisons of HF Across the Scenarios Using Post-Hoc Analysis From the Friedman Test.

Pairwise Comparisons	Std. Test Statistic	Sig.	Adj. Sig. (p)	r
HF_S1-HF_S3	−0.162	.871	1.000	.04
HF_S1-HF_S2	−0.324	.746	1.000	.07
HF_S3-HF_S2	0.162	.871	1.000	.04

In the case of RMSSD, significant differences were found between S2 and S1 (adj. P < .001, r = .93) and between S3 and S1 (adj. P = .004, r = .74), while the difference between S2 and S3 was not significant (adj. P = 1.000, r = .19; see Table 10). These results support the earlier findings that higher workload intensity corresponds to increased physiological stress, reflected in reduced RMSSD values. Although S3 was adjusted to each participant’s individual capacity and showed higher RMSSD values than S2, the absence of a significant difference between the two scenarios suggests only partial stress reduction. However, the rank distribution (Figure 6(c)) shows that in S3, fewer participants were assigned the lowest rank (1) than in S2, while more participants were assigned the middle rank (2), indicating subtle signs of stress mitigation.

Table 10.

Pairwise Comparisons of RMSSD Across the Scenarios Using Post-Hoc Analysis From the Friedman Test.

Pairwise Comparisons	Std. Test Statistic	Sig.	Adj. Sig. (p)	r
RMSSD_S2-RMSSD_S3	−0.811	.417	1.000	.19
RMSSD_S2-RMSSD_S1	4.056	<.001	.000	.93
RMSSD_S3-RMSSD_S1	3.244	.001	.004	.74

For AT, two out of the three pairwise comparisons were statistically significant (Table 11). Significant differences were found between S1 and S2 (adj. P < .001, r = 1.00) and between S1 and S3 (adj. P < .001, r = .93), indicating that task efficiency improved when the workload intensity increased or was personalized to individual capacity. The comparison between S2 and S3 was not significant (adj. P = .768, r = .26), showing that performance remained at a similarly high level under the personalized condition.

Table 11.

Pairwise Comparisons of AT Across the Scenarios Using Post-Hoc Analysis From the Friedman Test.

Pairwise Comparisons	Std. Test Statistic	Sig.	Adj. Sig. (p)	r
AT _S2-AT _S3	−1.136	.256	.768	.26
AT _S2-AT _S1	5.191	<.001	.000	1.00
AT _S3-AT _S1	4.056	<.001	.000	.93

Discussion

The purpose of the experimental design in E2 was to examine whether adjusting workload intensity to individual capacity would reduce physiological strain and stress response while maintaining task efficiency. The hypotheses for E2 predicted that the three scenarios would differ in their physiological and performance outcomes, with S3 expected to offer a more balanced response than the extremes represented by S1 and S2.

The physiological and performance indicators collectively support this interpretation. The Friedman test confirmed significant differences in HR and RMSSD across conditions, and post-hoc comparisons showed that S2 differed significantly from both S1 and S3. These findings reinforce the pattern observed in E1: elevated robot pacing increases sympathetic activation and reduces parasympathetic modulation, indicating a higher stress response. The magnitude of these differences was substantial, with HR showing a significant change between S1 and S2 and a significant reduction from S2 to S3. Additionally, the differences in RMSSD between S1 and S2 were also substantial. However, the smaller effect between S2 and S3 indicates that the personalized condition supported only partial recovery of parasympathetic activity. Although RMSSD did not differ significantly between S2 and S3 in pairwise testing, both the descriptive values and the distribution of ranks show a meaningful shift away from the lowest values recorded in S2, suggesting moderate stress mitigation under personalized conditions. HF remained unchanged across all scenarios, consistent with the stability observed in E1, indicating that HF is less sensitive to short-term workload fluctuations in HRC.

Regarding performance, the comparison of ATs between S1 and both higher-intensity scenarios (S2 and S3) confirmed significantly faster task execution under increased or personalized workload conditions. The corresponding effects were substantial, indicating that workload intensity had a strong and consistent impact on performance. The lack of a significant difference between S2 and S3 further suggests that high performance levels were maintained even when the workload was tailored to individual capacities, confirming that a human-centric approach (S3) can preserve productivity while reducing physiological stress.

Taken together, these findings support the rejection of the null hypotheses for HR, RMSSD, and AT in E2, providing evidence that the scenarios differ meaningfully in their impact on physiological strain and task efficiency. The null hypothesis for HF is retained, indicating that this measure did not vary in response to the workload manipulations used in the experiment. Overall, the results of E2 indicate that individualized workload intensity provides a more balanced trade-off between performance and well-being compared to either low- or high-intensity conditions alone. S3 successfully mitigated stress responses relative to S2, most clearly seen in HR reductions and modest RMSSD recovery, while maintaining task efficiency.

These findings provide empirical support for developing adaptive collaborative systems that dynamically adjust robot pacing based on individual worker capacity, rather than imposing uniform speed settings on all operators.

General Discussion and Conclusion

This study examined how varying levels of workload intensity influence workers’ well-being (specifically physiological stress responses) and task efficiency in HRC, and whether aligning robot pacing with individual worker capacity can provide a more sustainable balance between productivity and well-being. The research incorporated two sequential experimental structures (E1 and E2), which enabled an initial observation of the consequences of nonpersonalized workload extremes, followed by an assessment of a personalized, capacity-based alternative designed to mitigate physiological strain while maintaining high performance.

In E1, participants exhibited significant increases in HR and decreases in RMSSD under high-intensity conditions (S2), which are indicative of an elevated physiological stress response. These changes were not only statistically significant but also substantial in magnitude, reflecting large effects in both markers of autonomic activation. At the same time, performance, measured through AT, improved, suggesting that higher workload intensity can boost efficiency, but at the cost of increased strain and physiological stress response. HF, a marker of parasympathetic activity, remained largely unchanged, indicating its limited sensitivity to short-term workload variations in this context. Overall, the E1 findings reveal a classic trade-off: performance improves at high workload intensity, but at the expense of elevated physiological stress response.

To address this imbalance, in E2, an S3 was introduced, designed to adjust the CR motion parameters and their correlated workload intensity according to each participant’s individual capacity. The results showed that S3 maintained high efficiency successfully (comparable AT to S2), while reducing the physiological stress response (lower HR and higher RMSSD than in S2). The improvements in HR and RMSSD were supported by meaningful effect sizes, with HR showing a notably strong reduction and RMSSD showing a smaller but still meaningful recovery under personalized conditions. Notably, HF remained stable across all three scenarios, reinforcing the assumption that it may be less reactive to acute task demand changes or more influenced by individual variability. It is also possible that the sample size was too small to detect changes in the HF, or that the exposure to the experimental conditions was too short to induce changes in the HF.

Together, these findings provide two key insights. First, the benefits of high-intensity collaboration come at a physiological cost, raising concerns about the sustainability of such conditions in real industrial environments, where workers may be exposed to elevated intensity for prolonged periods. Second, adjusting the workload level, or personalizing robot pacing to match individual capacity, offers a balanced alternative. S3 successfully avoided the underload observed in S1 and the overload present in S2, providing a working tempo that sustained efficiency while reducing strain. This demonstrates that human-centric HRC design is not inherently at odds with productivity; when aligned with human capacity, it can maintain high performance while preventing negative effects on human well-being.

The novel contribution of this study lies in its combined assessment of performance and autonomic stress markers across low, high, and personalized workload conditions in a controlled HRC task. Few existing studies manipulate workload intensity parametrically, and even fewer evaluate personalization based on individual execution capacity (Javernik et al., 2023). The present findings, therefore, extend the current HRC literature by providing quantitative evidence that personalized workload intensity is not merely desirable but functionally effective (Borghi et al., 2024; Panchetti et al., 2023). Moreover, the study offers a structured experimental framework for evaluating adaptive collaborative systems and provides a replicable method for determining individual capacity and translating it into CR motion parameters.

Several limitations should be acknowledged. The sample size in E2 was relatively small, which limits broader generalization. Additionally, the scenario order was fixed, meaning that learning, fatigue, or anticipation effects cannot be fully ruled out. Although ECG recordings were visually inspected and artifact-free epochs were extracted, reliance on a single ten-minute window per scenario may limit insights into temporal dynamics. The stability of HF may reflect insensitivity to the task conditions but could also stem from short exposure durations or respiration-related variability. Finally, the experimental task was a simplified assembly operation, and more complex or cognitively demanding tasks may elicit different patterns of stress and performance across varying workload intensities.

Future research should explore real-time adaptive robot control based on continuous physiological monitoring, test adaptive pacing in realistic industrial environments, and include more diverse participant samples spanning a wider age range. Studies employing randomized scenario order and longer task durations will help determine whether the observed patterns persist under real-world temporal dynamics. Longitudinal investigations would also be valuable to understand better the implications of personalized HRC design for long-term health, fatigue accumulation, and job satisfaction.

In conclusion, this study demonstrates that workload intensity plays a decisive role in shaping physiological stress responses and performance in HRC. While high-intensity collaboration enhances efficiency, it simultaneously elevates physiological strain. A personalized approach to workload intensity, aligned with individual worker capacity, maintains high productivity while reducing stress. These conclusions are supported by consistent effect-size patterns across both experiments, highlighting that personalization offers a meaningful pathway toward more balanced and human-centric collaborative systems. These findings provide actionable guidance for designing adaptive, human-centric HRC systems in which worker well-being and operational performance are jointly optimized.

Key Points

Higher workload intensity in HRC increased HR and decreased RMSSD, indicating elevated physiological stress.

Assembly times improved under high-intensity conditions, but at the cost of reduced well-being.

A personalized, human-centric scenario (S3) maintained efficiency while lowering stress compared to the high-intensity condition.

The findings demonstrate the value of adaptive, human-centric HRC design that aligns robot pacing with individual worker capacity to optimize productivity and worker health jointly.

Supplemental Material

Supplemental Material - Material for Workload Intensity in Human-Robot Collaboration: A Personalized Approach to Sustainable Efficiency and Well-Being

Supplemental Material for Workload Intensity in Human-Robot Collaboration: A Personalized Approach to Sustainable Efficiency and Well-Being by Aljaz Javernik, Robert Ojstersek, and Katja Kerman in Human Factors.

Footnotes

Acknowledgments

The authors acknowledge the use of research equipment system for development and testing cognitive production approaches in industry 4.0: Collaborative robots with equipment and Sensors, hardware and software for ergonomic analysis of a collaborative workplace, procured within the project “Upgrading national research infrastructures - RIUM,” which was cofinanced by the Republic of Slovenia, the Ministry of Education, Science and Sport, and the European Union from the European Regional Development Fund.

Author Contributions

AJ, RO, and KK designed the study. AJ and RO collected the data. AJ and KK analyzed the data. All the authors drafted the manuscript.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The authors gratefully acknowledge the support of the Slovenian Research Agency (ARRS), Research Core Funding No. P2-0190, and the University of Maribor under the Internal Call for Financial Incentives for the Development of Research Activities in 2024, Agreement No. 075/2024/42-SPO.

Ethical Considerations

This study protocol was approved by the Research Ethics Committee of the Faculty of Arts, University of Maribor (038-03-169/2024/6/FF/UM).

ORCID iD

Aljaz Javernik

Supplemental Material

Supplemental material for this article is available online.

Author Biographies

Aljaž Javernik is a PhD student at the Faculty of Mechanical Engineering, University of Maribor, in the Laboratory for Production and Operations Management. He received his MSc in mechanical engineering from the University of Maribor in 2021.

Robert Ojsteršek is an assistant professor at the Faculty of Mechanical Engineering, University of Maribor, in the Laboratory for Production and Operations Management. He received his PhD in mechanical engineering from the University of Maribor in 2020.

Katja Kerman is an assistant professor in the Department of Psychology at the Faculty of Arts, University of Maribor, and the Department of Psychology, Faculty of Medicine, University of Maribor, Slovenia. She received her PhD in work and organizational psychology from the University of Vienna in 2021.

References

Baevsky

R. M.

Berseneva

A. P.

(2008). Use KARDIVAR system for determination of the stress level and estimation of the body adaptability.

Borghi

Ruo

Sabattini

Peruzzini

Villani

(2024). Assessing operator stress in collaborative robotics: A multimodal approach. Applied Ergonomics, 123(1), Article 104418. https://doi.org/10.1016/j.apergo.2024.104418

Bouaziz

Bettayeb

Sahnoun

Yassine

(2024). Incorporating uncertain human behavior in production scheduling for enhanced productivity in Industry 5.0 context. International Journal of Production Economics, 274(1), Article 109311. https://doi.org/10.1016/j.ijpe.2024.109311

Breque

De Nul

Petridis

(2021). Industry 5.0: Towards a sustainable, human-centric and resilient European industry. European Commission.

Calzavara

Faccio

Granata

Trevisani

(2024). Achieving productivity and operator well-being: A dynamic task allocation strategy for collaborative assembly systems in Industry 5.0. International Journal of Advanced Manufacturing Technology, 134(7–8), 3201–3216. https://doi.org/10.1007/s00170-024-12334-9

Cohen

(1992). A power primer. Psychological Bulletin, 112(1), 155–159. https://doi.org/10.1037/0033-2909.112.1.155

Cohen

(2013). Statistical power analysis for the behavioral sciences (2nd ed.). Taylor & Francis.

Coronado

Kiyokawa

Ricardez

G. A. G.

Ramirez-Alpizar

I. G.

Venture

Yamanobe

(2022). Evaluating quality in human–robot interaction: A systematic search and classification of performance and human-centered factors, measures and metrics towards an Industry 5.0. Journal of Manufacturing Systems, 63(9), 392–410. https://doi.org/10.1016/j.jmsy.2022.01.006

Friedman

(1937). The use of ranks to avoid the assumption of normality implicit in the analysis of variance. Journal of the American Statistical Association, 32(200), 675–701. https://doi.org/10.1080/01621459.1937.10503522

10.

Halkos

Bousinakis

(2010). The effect of stress and satisfaction on productivity. International Journal of Productivity and Performance Management, 59(5), 415–431. https://doi.org/10.1108/17410401011052869

11.

Hart

S. G.

Staveland

L. E.

(1988). Development of NASA-TLX (Task Load Index): Results of empirical and theoretical research. In Hancock

P. A.

Meshkati

(Eds.), Advances in psychology (Vol. 52, pp. 139–183). North-Holland.

12.

Hopko

Wang

Mehta

(2022). Human factors considerations and metrics in shared space human–robot collaboration: A systematic review. Frontiers in Robotics and AI, 9(1), Article 799522. https://doi.org/10.3389/frobt.2022.799522

13.

Javernik

Buchmeister

Ojsteršek

(2023). The NASA-TLX approach to understand workers’ workload in human–robot collaboration. International Journal of Simulation Modelling, 22(4), 574–585. https://doi.org/10.2507/IJSIMM22-4-673

14.

Montini

Cutrona

Dell’Oca

Bettoni

Landolfi

Rocco

Carpanzano

(2024). An industrial human–robot collaboration case study for workers’ well-being. Procedia CIRP, 130(1), 924–929. https://doi.org/10.1016/j.procir.2023.12.155

15.

Ojsteršek

Javernik

Buchmeister

(2024). Integrating simulation modelling for sustainable, human-centred Industry 5.0: ESG-based evaluation in collaborative workplaces. Advances in Production Engineering & Management, 19(4), 527–538. https://doi.org/10.14743/apem2024.4.496

16.

Panagou

Neumann

W. P.

Fruggiero

(2024). A scoping review of human robot interaction research towards Industry 5.0 human-centric workplaces. International Journal of Production Research, 62(3), 974–990. https://doi.org/10.1080/00207543.2022.2146235

17.

Panchetti

Pietrantoni

Puzzo

Gualtieri

Fraboni

(2023). Assessing the relationship between cognitive workload, workstation design, user acceptance and trust in collaborative robots. Applied Sciences, 13(3), 1720. https://doi.org/10.3390/app13031720

18.

Shaffer

Ginsberg

J. P.

(2017). An overview of heart rate variability metrics and norms. Frontiers in Public Health, 5(1), 258. https://doi.org/10.3389/fpubh.2017.00258

19.

Sharma

Gedeon

(2012). Objective measures, sensors and computational techniques for stress recognition and classification: A survey. Computer Methods and Programs in Biomedicine, 108(3), 1287–1301. https://doi.org/10.1016/j.cmpb.2012.07.003

20.

Simões

A. C.

Pinto

Santos

Pinheiro

Romero

(2022). Designing human–robot collaboration (HRC) workspaces in industrial settings: A systematic literature review. Journal of Manufacturing Systems, 62(94), 28–43. https://doi.org/10.1016/j.jmsy.2021.11.006

21.

Feng

Fan

(2024). Exploring the impact of human–robot interaction on workers’ mental stress in collaborative assembly tasks. Applied Ergonomics, 116(1), Article 104224. https://doi.org/10.1016/j.apergo.2023.104224

22.

Tarvainen

M. P.

Niskanen

J. P.

Lipponen

J. A.

Ranta-Aho

P. O.

Karjalainen

P. A.

(2014). Kubios HRV–Heart rate variability analysis software. Computer Methods and Programs in Biomedicine, 113(1), 210–220. https://doi.org/10.1016/j.cmpb.2013.07.024

23.

van Dijk

Baltrusch

S. J.

Dessers

de Looze

M. P.

(2023). The effect of human autonomy and robot work pace on perceived workload in human–robot collaborative assembly work. Frontiers in Robotics and AI, 10(1), Article 1244656. https://doi.org/10.3389/frobt.2023.1244656

24.

Villani

Pini

Leali

Secchi

(2018). Survey on human–robot collaboration in industrial settings: Safety, intuitive interfaces and applications. Mechatronics, 55(1), 248–266. https://doi.org/10.1016/j.mechatronics.2018.02.009

25.

World Health Organization . (2022). WHO guidelines on mental health at work. WHO.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.07 MB

Workload Intensity in Human–Robot Collaboration: A Personalized Approach to Sustainable Efficiency and Well-Being

Abstract

Objective

Background

Methods

Results

Conclusions

Application

Keywords

Introduction

Purpose of Research

Methods

Independent Variables

Dependent Variables

Experiment 1

Participants

Human–Robot Collaboration Setup and Assembly Operation

Experimental Design

Statistical Methods

Results

Discussion

Experiment 2

Participants

Experimental Design

Statistical Methods

Results

Discussion

General Discussion and Conclusion

Key Points

Supplemental Material

Supplemental Material - Material for Workload Intensity in Human-Robot Collaboration: A Personalized Approach to Sustainable Efficiency and Well-Being

Footnotes

Acknowledgments

Author Contributions

Declaration of Conflicting Interests

Funding

Ethical Considerations

ORCID iD

Supplemental Material

Author Biographies

References

Supplementary Material