Evaluating Human Potential to Control Multiple Degrees of Freedom via User-Defined Lower-Limbs Movements

Evaluating Human Potential to Control Multiple Degrees of Freedom via User-Defined Lower-Limbs Movements Elena Bastianelli University of Siena Siena, Italy Tommaso Lisini Baldi University of Siena Siena, Italy Nicole D’Aurizio University of Siena Siena, Italy Elena De Momi Politecnico di Milano Milan, Italy Domenico Prattichizzo University of Siena Siena, Italy Abstract—In human augmentation, researchers have successfully demonstrated the control of supernumerary limbs but none have investigated the human potential to control the maximum number of degrees of freedom (DoFs) from both kinematic and cognitive perspectives. Existing works impose both the body part to be used and the specific movement to perform, limiting user adaptability. Here, in order to characterize the human potential, we implemented a strategy enabling the control of multiple DoFs through userdefined movements. Subjects can efficiently manage up to two DoFs with little practice, good performance, and acceptable perceived workload. Consistent performance when controlling more than two DoFs requires additional training. Index Terms—human augmentation, human-centred robotics I. INTRODUCTION In today’s world, people spend between 9 and 12 hours per day seated, amounting to 65–80 % of their time. Because of modern lifestyle, people are typically seated during the most of their tasks. This may be a consequence of human evolution, aiming to save time and energy and leading to technologies, such as autonomous robots or artificial intelligences, that can replace humans. Differently, human augmentation aims at making the most of our time enhancing human input to support people rather than replace them. Indeed, exploiting human body’s potential to integrate augmentation systems into work environments could allow operators to manage robotic tools or interfaces while remaining seated and keeping their hands free. The research is paving the way for applications where seated individuals can control robotic systems using their legs. Since all existing solutions impose the gesture and the DoF to control [1], this study aims to assess the feasibility of controlling up to four DoFs by freely choosing one’s own lower-body movements and their corresponding DoF, in terms of both performance and workload. II. METRICS OF INTEREST The Fitts’ Law, which models the human movement in a pointing task, is one of the best-know attempt to mathematically describe the speed-accuracy tradeoff, e.g., slowing down a movement to be more precise. The movement time MT to reach a target is: MT =aID +b where ID , the index of difficulty of the target, expresses how the time depends on the distance to the target D and its width W: ID = log2D W+ 1[bit]. Because the intercept b can arise just as an adjustment between two equivalent IDs [2], the slope a corresponds to the performance index. NASA Task Load Index (NASA-TLX) [3] was used to evaluate the workload through subjects’ subjective scales of performance. III. EXPERIMENTS Participants, tracked through 7 IMU sensors (Awinda Xsens, NL), aligned a virtual cursor with multiple targets, controlling a different number of DoFs with their respective gestures. These were acquired in the calibration phase, where subjects could chose gestures that felt intuitive for the corresponding DoF. All of the eight participants performed the following experiments: Experiment 1 - One DoF: the subject aligned the x -position of the cursor with the one of target. Experiment 2 - Two DoFs: the participant matched both the x - axis and the y -axis of the cursor with the xy -coordinates of the target. Experiment 3 - Three DoFs: the subject controlled both the 2D position of the cursor and its rotation around the z-axis. Experiment 4 - Four DoFs: subject also controlled the diameter of the cursor to match the the one of the target, alongside its xy-position and its rotation along the z-axis. Each experiment was composed of 3 blocks of 12 targets (144 targets per person across the entire campaign: 4 experiments × 3 blocks × 12 targets). Only one target appeared at a time in the scene. At the beginning of the trial and after each target, the subject was asked to come back to the starting position. At the end of each experiment participants filled in the NASATLX questionnaire. Only after the first experiment and before rating the single factors, participants were asked to perform a pairwise comparison between the sources of workload, indicating which factors contributed most to the perceived workload when performing the pointing task. In this way, each factor is associated with a certain weight, and the overall workload is computed as the weighted average of all the ratings. IV. RESULTS A linear mixed-effects model was fitted to investigate the effects of both the number of DoFs and the ID on MT . Both number of DoFs ( p < 0.001 ) and ID ( p= 0.025 ) had statistically significant effects. Results are reported in Fig. 1. A one-way repeated measures ANOVA was conducted to evaluate the impact of the number of DoFs on the perceived workload. Perceived workload increased from 28.9±7.5 for 1 DoF, to 43.5±7.1 for 2 DoFs, to 57.3±6.7 for 3 DoFs, and to 68.0±5.5 for 4 DoFs. The post hoc analysis with a Bonferroni adjustment indicated a statistically significant increase of 39.1 only in perceived workload from 1 DoF to 4 DoFs ( p= 0.016 ). 2025 I-RIM Conference October 17-19, Rome, Italy ISBN: 9788894580570 10.5281/zenodo.17629648 65 0.511.522.533.544.5 0 5 10 15 20 25 30 35 ID [bit] MT [s] (a) 1234567 0 5 10 15 20 25 30 35 ID [bit] MT [s] (b) 23456789 0 5 10 15 20 25 30 35 ID [bit] MT [s] (c) 0 2 4 6 8 10 12 5 10 15 20 25 30 35 ID [bit] MT [s] (d) Fig. 1: Experimental results on performance. Plots display the collected data and the regression lines across the four experimental conditions. Solid lines represent the main effect of average performance, while dashed lines represent best performance. In a) data from Experiment 1, in b) data from Experiment 2, in c) data from Experiment 3, in d) data from Experiment 4. Effort Frustration Mental Demand Performance Physical Demand Temporal Demand 0 20 40 60 80 100 ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ 1 DoF 2 DoFs 3 DoFs 4 DoFs (a) 1 DoF 2 DoFs 3 DoFs 4 DoFs 0 20 40 60 80 100 ∗ (b) Fig. 2: In a),the average ratings among subjects for the different factors (range: 0-100) are reported in red for 1 DoF, in yellow for 2 DoFs, in blue for 3 DoFs, and in green for 4 DoFs. In b), the overall workload (i.e., the mean of weighted ratings). Lower ratings indicate a better outcome. Statistical significance of differences is denoted as ∗for p < 0.05. A deeper analysis of the single factors was conducted to gain further insights on the results, as shown in Fig. 2. V. DISCUSSION For 1 and 2 DoFs control, it was easier for subjects to achieve high performance, suggesting a lower entry barrier for mastering these tasks. Results for 3 and 4 DoFs exhibit significantly higher variability compared to the first two conditions, indicating that their results may not be entirely reliable. The intercept represents a temporal bias reflecting the starting point for the subject in controlling a certain number of DoFs. It is obvious that managing sequential multiple-DoFs requires more time as the number of DoFs increases. The technology may also require more time to discriminate between more control gestures. Indeed, the intercept reflects the system minimal time delay, considering both the subject’s behavior and the movements selected. Movements that are either too similar or unintuitive can introduce computational delays and mental confusion. The high variability of inputs lifted the observations variance but this was necessary to model a complete set of individual performances and retrieve the human potential, i.e., the optimal number of controllable DoFs. In this sense, the proposed method proved to be reliable, enabling people to control multiple DoFs. VI. CONCLUSIONS This research identifies the human-potential in controlling multiple DoFs with the lower-limbs in a seated posture, in terms of performance and workload. It also validates the innovative human augmentation strategy enabling control of up to four DoFs using arbitrary lower-body movements. By comparing the control of 1 and 2 DoFs, we can conclude that these two conditions are equivalent across all measured aspects. This equivalence suggests that these strategies can be treated as comparable. For the purpose of human augmentation, where maximizing subject participation and engagement is critical, we conclude that controlling 2 DoFs is the optimal balance. As regards controlling 3 or 4 DoFs, this work does not provide enough reliable data. We hypothesize that the increased difficulty of these tasks requires extensive training to help individuals become familiar with the technology, leading to more consistent performance. The novelty of our strategy lies in allowing people to freely choose how to control additional DoFs, and the present study demonstrates that it can be used for controlling multiple DoFs. Its adaptability to the individual enabled the assessment of each subject’s maximal potential, without ergonomic or anthropometric constraints limiting performance or increasing cognitive and physical workload. It has also been proved that the freedom in the choice of the movements may act as a doubleedged sword, likely due to the participants’ inexperience. While this flexibility enables subjects to select movements that feel more intuitive to them, it also reveals their lack of experience in identifying smoother, less strenuous, or more distinct movement patterns. REFERENCES [1] R. Raisamo et al., “Human augmentation: Past, present and future,” International Journal of Human-Computer Studies, vol. 131, 05 2019. [2] J. Gori et al., “Speed-accuracy tradeoff: A formal information-theoretic transmission scheme (fitts),” ACM Transactions on Computer-Human Interaction, vol. 25, pp. 1–33, 09 2018. [3] S. G. Hart et al., “Development of NASA-TLX (Task Load Index): Results of empirical and theoretical research,” in Advances in psychology. Elsevier, 1988, vol. 52, pp. 139–183. 66