JMIR Publications

ABSTRACT

Background:

With the rise of Industry 4.0, the integration of robots into industrial settings has rapidly advanced, aiming to reduce human involvement in repetitive and physically demanding tasks while improving overall efficiency. As collaborative robots become more prevalent, assessing the physical strain experienced by human workers during joint tasks is essential – particularly from an ergonomic perspective – to ensure safety, prevent musculoskeletal disorders, and promote long-term well-being in the workplace.

Objective:

To investigate how Human-Robot Collaboration during manipulation tasks performed in parallel task configurations while standing influences workers’ postural control and musculoskeletal load.

Methods:

Fourteen healthy male subjects performed specifically designed manipulation tasks under three different experimental conditions: without robotic assistance (Free - F), with a freely moving cobot providing load support (Robot Free - RF), and with a cobot constrained to horizontal movement only (Robot Plane - RP), all involving a predefined object manipulation sequence. Center of pressure (COP) trajectories were extracted from two force plates data to calculate traditional center of pressure linear metrics (mean distance, mean velocity, confidence ellipse area, and sway area) and non-linear Recurrence Quantification Analysis (RQA) indicators (recurrence, determinism, and ratio between them).

Results:

Statistical analysis of the four COP-derived parameters showed significantly greater postural sway in both robot-assisted conditions compared to the F condition, with higher values in mean distance (approximately 2.2 cm in RF and RP vs. 1.5 cm in F), mean velocity (approximately 4.2 cm/s in RF and RP vs. 3 cm/s in F), confidence ellipse area (approximately 20 cm2 in RF and RP vs. 8 cm2 in F), and sway area (approximately 3 cm2/s in RF and RP vs. 1.5 cm2/s in F). Statistical analysis on the RQA metrics revealed a lower recurrence rate in robot-assisted conditions (approximately 0.2 in RF and RP vs. 0.3 in F) but stable determinism (values close to 1), leading to a higher determinism-to-recurrence ratio. No significant differences were found between the two robot-assisted conditions in any analyzed parameter.

Conclusions:

Interacting with the cobot leads to greater and more extensive postural sway, indicating reduced stability and higher physical demand compared to performing the task alone. This could reflect either impaired balance or adaptive motor responses due to the cobot influence on the body internal representation. RQA analysis reveals that, while cobot interaction disrupts postural consistency and increases variability, it maintains a structured control of balance, which seems to require continuous adjustments rather than random instability. The lack of significant differences between the two robot-assisted modes suggests that the mere presence of the cobot – rather than its movement constraints – is the primary driver of these postural changes.