Fig 1.
(A) Environment: 9 Nexus Vicon infrared cameras (orange circles) and a custom-made library made of a table and a shelf. (B) Positioning of IR-reflective markers on subjects (1)C7, (2)T10, (3)Scapula, (4)Shoulder, (5)Arm, (6)Elbow, (7)Forearm, (8)Outer Wrist, (9)Inner Wrist, (10) Collar bone, (11)Sternum, (12)Knuckle (C) Proposed pick-and-place task: 3 movements from the desk to the shelf (M1-M3), 3 from shelf to desk (M4-M6), and 3 from desk to desk (M7-M9).
Fig 2.
Physical set-up and a participant performing the proposed pick-and-place task.
Table 1.
Labels for the proposed 9 pick-and-place movements.
Movements performed in the same direction where clustered together and labeled under one name.
Fig 3.
Representations of absolute hand positions (A) and velocities (B) across each subject, repetition, and movement.
Each curve was normalized over their own duration thus velocity is expressed as m/path. M1-M3 can be grouped as g-, M4-M6 as g+, and M7-M9 as g0.
Table 2.
Seven kinematic features were selected according to 4 different performance evaluations: Smoothness, planning, efficiency, and accuracy.
Abbreviations are reported, as well as literature references, if present.
Fig 4.
Representation of velocity peaks, number of peaks in the velocity profile and Movement Time distribution for each gravity condition.
(A) Each row represents the absolute hand velocities for either g−, g+, and g0 sets of movements. The maximum peak is marked in red meanwhile all the local maxima are marked as blue stars. (B) Distribution of number of velocity profiles presenting 1, 2, 3, or 4 local maxima. (C) Movement Time (MT) distribution for each gravity condition. The arrangement is positively skewed for g− (1.43), g+ (1.77), and g0 (1.44).
Table 3.
Kinematic metrics calculated from 24 healthy subjects while performing 9 pick-and-places movements.
Fig 5.
Statistical distribution of kinematic metrics.
Graphs A, B, and C show the statistical distribution of PTPV, PTPSD, and SPARC metrics respectively. Each metric was calculated for three sets of movements (i.e. g-, in pink, g+, in violet, g0, in green). From these graphs it’s possible to notice how the distributions for PTPV, PTPSD, and SPARC metrics are comparable between g− and g0, as the means are mostly aligned with respect to g+. Graph D reports the comparison between PTPV (in red) and PTPSD (in blue) for each gravity condition. Overall, maximum peak is reached before maximum standard deviation across subjects, trials, and movements.
Fig 6.
Representation of Target Position Error and Minimum Required Tunnel metrics.
(A) 3D and 2D view of Target Position Error (TPE) metric defined as the sphere whose radius contains up to 95% of hand end-points. Sphere related to g− (in pink) and g+ (in violet) are almost symmetrical and present radius of 3.33 and 3cm respectively. g0 sphere (in green) radius is 2.3cm. The small dots represent experimental data, each colored in accordance to the gravity condition they belong to. (B) Representation of the average standard deviations across subjects, trials and movements for g−, g+, and g0. Solid line is representative of 1 standard deviation, meanwhile dashed line of 3 standard deviations. The Minimum Required Tunnel (MRT) metric is defined as the integral between solid and dashed lines (Table 3), shaded in the graph.
Table 4.
Statistical analysis on the distribution of each set of movement (g−, g+, g0).
Repeated measures ANOVA analysis was performed for NVP, PTPV, SPARC, and PTPSD metrics. Friedman test was used for NVP (α = 0.05). Pair-wise Post-hoc t-student test were conducted for PTPV, PTPSD, and SPARC metric, post-hoc Mann-Whitney test for MT (Bonferroni correction, α = 0.017).