Fitts' List Revisited: An Empirical Study on Function Allocation in a Two-Agent Physical Human-Robot Collaborative Position/Force Task

TL;DR

The paper investigates whether Fitts' MABA-MABA principle applies to static function allocation in two-agent physical human–robot collaboration (pHRC) by evaluating four position/force allocation conditions in a blending task with 26 participants. It combines objective performance metrics (completion time, overblending) with multi-dimensional subjective measures (Van der Laan acceptance, NASA-TLX, ACES) to assess user experience. The results show that human control of position with robot control of force (HR) yields better acceptance and lower overblending, while fully autonomous supervision (RR) provides fastest completion and lowest workload at the cost of perceived autonomy; robot-controlled position with human force (RH) tends to produce higher frustration. Overall, the findings validate applying Fitts' allocation principles to pHRC while highlighting nuanced UX trade-offs, particularly regarding autonomy and engagement in shared physical tasks, with implications for Industry 5.0 deployment.

Abstract

In this letter, we investigate whether classical function allocation-the principle of assigning tasks to either a human or a machine-holds for physical Human-Robot Collaboration, which is important for providing insights for Industry 5.0 to guide how to best augment rather than replace workers. This study empirically tests the applicability of Fitts' List within physical Human-Robot Collaboration, by conducting a user study (N=26, within-subject design) to evaluate four distinct allocations of position/force control between human and robot in an abstract blending task. We hypothesize that the function in which humans control the position achieves better performance and receives higher user ratings. When allocating position control to the human and force control to the robot, compared to the opposite case, we observed a significant improvement in preventing overblending. This was also perceived better in terms of physical demand and overall system acceptance, while participants experienced greater autonomy, more engagement and less frustration. An interesting insight was that the supervisory role (when the robot controls both position and force) was rated second best in terms of subjective acceptance. Another surprising insight was that if position control was delegated to the robot, the participants perceived much lower autonomy than when the force control was delegated to the robot. These findings empirically support applying Fitts' principles to static function allocation for physical collaboration, while also revealing important nuanced user experience trade-offs, particularly regarding perceived autonomy when delegating position control.

Figures (6)

• Figure 1: Illustrative comparison of a manual blending task (left) versus a collaborative robot assisting a human worker (right) by potentially handling force or position control, based on function allocation.
• Figure 2: Illustration of pipeline for the method. The experimental setup on the left involves a blending task with force (red) and position (green) sub-tasks. The combinations of assigning these sub-tasks between human (H) and robot (R) result in four experimental conditions. In the labeling of conditions, the first letter is for the position sub-task, while the second letter is for the force sub-task (e.g., the human doing positioning and the robot producing force would be HR). The experiment procedure is followed by the calculation of measures, which are finally used in the statistical evaluation.
• Figure 3: Plot showing subjective scores on the Van der Laan acceptance scales: usefulness (y‑axis) versus satisfying (x‑axis). Large circles with error bars show the mean $\pm$ SD for each condition, while small semi‑transparent dots are the corresponding single‑participant scores. Points that lie higher and further right indicate greater perceived acceptance. Statistically significant pair‑wise differences are annotated on each axis, where: **: $0.001 < p \le 0.01$; ***: $0.0001 < p \le 0.001$; ****: $p \le 0.0001$.
• Figure 4: Box plots showing NASA‑TLX scores. Each swimlane shows the results of one workload sub‑scale for each of the conditions. Boxes show the median and the interquartile range, where whiskers extend to the 5th and 95th percentiles, while small dots are the individual participant scores. Higher scores correspond to greater perceived workload. Statistically significant pair‑wise differences are annotated, where: *: $0.01 < p \le 0.05$; **: $0.001 < p \le 0.01$; ***: $0.0001 < p \le 0.001$; ****: $p \le 0.0001$.
• Figure 5: Box‑plots showing objective performance metrics in terms of completion time in seconds (top) and cumulative overblended health in health points (bottom). Boxes show the median and the interquartile range, where whiskers extend to the 5th and 95th percentiles, while each small dot is one trial and is colored according to the trial number. Lower values correspond to better performance. Statistically significant pair‑wise differences are annotated, where: *: $0.01 < p \le 0.05$; ****: $p \le 0.0001$.