A dataset of wrist reaching movements for the study of collaborative human motor control

Motivation Understanding human-human collaboration during goal-directed movements could allow to guide the development of interactive controllers for contact robots (Li et al. 2022), including exoskeletons (Dalla Gasperina et al. 2021), supernumerary robotic limbs (Eden et al. 2022) and cobots (Liu et al. 2022). Although multiple results have been obtained regarding the improved performance of humans collaboratively tracking targets (Takagi et al. 2017), goal-directed movements remain understudied. To our knowledge, there is no publicly available dataset allowing to systematically study such collaborative movements, performed with different interaction and movement dynamics. Short description This dataset includes wrist movement kinematics and human-human interaction forces from 20 participants (10 dyads), collected through dedicated active exoskeletons (HRX-1, Human Robotix, London). General anthropometric data from the participants, and the order in which they performed the different conditions, are available in the file "Participants_data.csv". Methodology Participants were asked to perform horizontal wrist flexion/extension movements while connected to the exoskeletons, to move a cursor towards targets displayed on a screen. Five movements amplitude were tested, both alone and with a high and low human-human connection stiffness (spring-like connection). Furthermore, a viscous force field, with either high or low viscosity, was added to increase the cost of movement in several conditions. A detailed description of the methodology and saved parameters can be found in Verdel et al. (2025). Data description The main data folder "CoT_HumanHuman" contains two subfolders: (i) "raw_data" containing the raw outputs of the experiment, and (ii) "data' containing segmented movements and separators. Both subfolders are described below. Raw data This folder contains 10 subfolders, each corresponding to one of the human-human dyads and named "SiSj", where "i" and "j" are participants' identifiers. Each subfolder contains the following 11 files: JCoT_SiSj_Ctrl_1_1_mu_0_Kp_0_Kd_0_W_0.csv : Solo condition with passive exoskeletons; first experimental block JCoT_SiSj_Ctrl_1_2_mu_0_Kp_0_Kd_0_W_0.csv : Solo condition with passive exoskeletons; last experimental block JCoT_SiSj_Ctrl_1_mu_0_Kp_0_Kd_0_W_1.csv : Solo condition with passive exoskeletons (shortened for washout); fourth experimental block JCoT_SiSj_Ctrl_2_mu_0.15_Kp_0_Kd_0_W_0.csv : Solo condition with viscous exoskeletons; high viscosity (0.15 Nm s / rad) JCoT_SiSj_Ctrl_2_mu_0.075_Kp_0_Kd_0_W_0.csv : Solo condition with viscous exoskeletons; low viscosity (0.075 Nm s / rad) JCoT_SiSj_Ctrl_3_mu_0_Kp_0.5_Kd_0_W_0.csv : Connected condition with passive exoskeletons; low stiffness (0.5 Nm / rad) JCoT_SiSj_Ctrl_3_mu_0_Kp_1.6_Kd_0_W_0.csv : Connected condition with passive exoskeletons; high stiffness (1.6 Nm / rad) JCoT_SiSj_Ctrl_4_mu_0.15_Kp_0.5_Kd_0_W_0.csv : Connected condition with viscous exoskeletons; low stiffness (0.5 Nm / rad) and high viscosity (0.15 Nm s / rad) JCoT_SiSj_Ctrl_4_mu_0.15_Kp_1.6_Kd_0_W_0.csv : Connected condition with viscous exoskeletons; high stiffness (1.6 Nm / rad) and high viscosity (0.15 Nm s / rad) JCoT_SiSj_Ctrl_4_mu_0.075_Kp_0.5_Kd_0_W_0.csv : Connected condition with viscous exoskeletons; low stiffness (0.5 Nm / rad) and low viscosity (0.075 Nm s / rad) JCoT_SiSj_Ctrl_4_mu_0.075_Kp_1.6_Kd_0_W_0.csv : Connected condition with viscous exoskeletons; high stiffness (1.6 Nm / rad) and low viscosity (0.075 Nm s / rad) The columns in these ".csv" files correspond to the following variables: TIME, MVT_INDEX_Si, TARGET_Si, TARGET_POS_Si, CURSOR_POS_Si, WRIST_POS_Si, WRIST_VEL_Si, WRIST_TORQUE_Si, OVERSHOOT_Si, MVT_INDEX_Sj, TARGET_Sj, TARGET_POS_Sj, CURSOR_POS_Sj, WRIST_POS_Sj, WRIST_VEL_Sj, WRIST_TORQUE_Sj, OVERSHOOT_Sj Segmented data This folder contains 21 subfolders: (i) a "movements" folder containing the data of each movement performed by each participant, stored in a ".pickle" file, and (ii) a "Si" folder for each participant. In the "movements" folder, the files are named "rob_Si_COND_mk.pickle", where "i" $\in[\![1,20]\!]$ is the participant identifier, "COND" $\in\{$P1, P2, W, VL, VH, KL, KH, VL_KL, VL_KH, VH_KL, VH_KH$\}$ correspond to the different conditions investigated in Verdel et al. (2025), and "k" is the movement index in the considered condition. Note that the segmented movements stored in this folder contain raw position profiles and low-pass filtered velocity and acceleration profiles ($4^{th}$ order Butterworth, $5$ Hz cut-off), and low-pass filtered torque and interaction torque profiles ($4^{th}$ order Butterworth, $15$ Hz cut-off). The velocity and acceleration are obtained from the numerical differentiation of the position data through time. Note that ".pickle" files can be opened and processed using Python. In each "Si" folder, we provide: 11 files named "blocks_separators_Si_COND.csv": where the index of the first and last useful iterations in a condition are provided 11 files named "trials_separators_Si_COND.csv": where the index of the first and last iterations of each movement in a condition are provided 11 files named "Si_COND.csv": where the raw data of each participant is provided with headers and the interaction torque is added to the previous variables