Fig 1.
Experimental setup and centre of mass adaptation dynamics.
(A) Illustration of the experimental setup where participants perform squat‑to‑stand movements using a custom force‑controlled pulling mechanism to perturb their centre of mass. (B) Overview of the experimental protocol where light grey indicates unperturbed trials while dark grey indicates perturbed trials. (C) Failed and successful trajectories of a representative participant where dashed lines show baseline trajectory, red represents failed trials and grey represents successful trials. (D, E) Trial-by-trial progression of Trajectory Area and Initial Trajectory Area during the whole experiment where red represents failed trials, green represents successful trials, black represents all trials, dark blue represents the catch trial, and light blue represents the first de-adaptation trial. Error bars show standard error of the mean.
Fig 2.
Representative centre of mass trajectories and adaptation metrics.
(A) Representative set of centre of mass trajectories of two consecutive successful trials. (B, C) Total changes of Trajectory Area and Initial Trajectory Area, with changes between two consecutive successful trials (S-S) shown in pink and those between two consecutive successful trials separated by at least one failed trial (S-F-S) shown in blue. Error bars show standard error of the mean. (D) Representative set of centre of mass trajectories before and after a failed trial with successful trials shown in green and failed trials shown in red.
Fig 3.
Adaptation dynamics during perturbation and de-adaptation.
(A) The number of participants among the total of 20 that failed at each trial. (B, C) Progression of Trajectory Area and Initial Trajectory Area during adaptation to perturbation where red represents failed trials, green represents successful trials, and black represents all trials. (D) Representative set of centre of mass trajectories for unperturbed trial, for last perturbed trial, for catch trial and for first de-adaptation trial. (E, F) Comparison of Trajectory Area and Initial Trajectory Area of last perturbed trial, catch trial, and first de-adaptation trial. (G, H) Progression of Trajectory Area and Initial Trajectory Area during the de-adaptation block. Error bars show standard error of the mean.
Fig 4.
Progression of Smoothness and Co-Contraction during adaptation to perturbation.
(A, B) The progression of Smoothness and Co-contraction during adaptation to perturbation, with red representing failed trials, green representing successful trials, and black representing all trials. (C, D) Total changes of Smoothness and Co-Contraction where change between two consecutive successful trials (S-S) is shown in pink and change between two consecutive successful trials separated by at least one failed trial (S-F-S) is shown in blue. (E) Representative jerk of centre of mass trajectory and EMG trajectories for successful trials before a failed trial (light green) and after a failed trial (dark green). Error bars show standard error of the mean.
Fig 5.
Proposed Computational Model of Hierarchical Motor Learning.
The model integrates reinforcement learning, internal model adaptation, and feedback control to simulate human motor learning under ecological control. Reinforcement learning optimizes motor plans based on trial outcomes, the internal model adapts to new body-environment dynamics, and feedback control compensates for inaccuracies in inverse dynamics. The model follows the experimental protocol, enabling direct comparison with human data and capturing key aspects of adaptation, including trial-by-trial learning and the role of failures in shaping movement strategies.
Fig 6.
Simulation results of ecological and hierarchical motor learning model.
(A, B, C) Trajectory Area, Initial Trajectory Area, and Effort along normalised learning trials where red represents failed trials, green represents successful trials, and black represents all trials. (D) Representative set of simulated centre of mass trajectories for unperturbed trial, for last perturbed trial, for catch trial and for first de-adaptation trial. (E) Trajectory Area and Initial Trajectory Area changes over simulated experiments for 100 simulated participants for all trials, for two consecutive successful trials (S-S), and for consecutive successful trials separated by at least one failure (S-F-S). (F) Average number of failures among the simulated N = 59 participants at each trial number. (G) Average feedback gain multiplier as a function of perturbed trial number for failed, successful, and all trials.