Have a personal or library account? Click to login
Adaptive Upper Limb Robot-Assisted Rehabilitation: Learn-from-Therapist Demonstrations Cover

Adaptive Upper Limb Robot-Assisted Rehabilitation: Learn-from-Therapist Demonstrations

Journal of Automation, Mobile Robotics and Intelligent Systems
Volume 20 (2026): Issue 1 (March 2026)
By: Ismail Auta,  Ahmed Fares,  Hiroyasu Iwata and  Haitham El-Hussieny  
Open Access
|Mar 2026

Figures & Tables

Figure 1.

A block diagram of the proposed LfTD system, which trains a robot with two degrees of freedom to perform rehabilitation exercises. These exercises are learned from therapist demonstrations using DMP

Figure 2.

Schematic diagram of a two-dimensional DMP. This diagram illustrates how learned weights (ω) model trajectories, adjusting the motion from the start (y0) to the goal (g) positions

Figure 3.

Conceptual block diagram of the MRAC illustrating the adaptive control strategy by highlighting the feedback loop that dynamically adjusts controller gains (K, K˜\tilde K) to enhance disturbance robustness

Figure 4.

Two distinct demonstrations performed by human subject, depicting variations in movement patterns traced during demonstrations. (a) S-shaped pattern; (b) ’3’-shaped pattern

Figure 5.

Sixty equally spaced Gaussian distributions illustrating the basis functions used

Figure 6.

Performance optimization using PSO

Figure 7.

Reproduced demonstrated trajectory using the proposed LfTD technique

Figure 8.

Total position reproduction error in DMP modeling

Figure 9.

Adaptation of the demonstrated trajectory to changes in starting Points and goals. Figure (a) shows the trajectory adapted to a new goal, maintaining the original starting point. Figure (b) demonstrates how both the starting point and goal are changed

Figure 10.

LfTD framework in trajectory tracking performance between desired and actual trajectories. (a) Tracking for each DOF. (b) Tracking for the demonstrated trajectory executed by the robot

Figure 11.

Tracking the generated trajectory with added disturbances mimicking patient deviation for (a) each DOF and (b) for the demonstrated trajectory as executed by the robot

Parameters for the LfTD Model

ParametersDescriptionValue
nNo. of gaussians60
αDMP parameter8.5
βDMP parameter4
αxCanonical system constant0.8
τTime scaling constant0.218
rController adaptation rate100
KAdaptive controller gain[10 16]
liRobot link lengths[50, 50] cm

Mean Absolute Error (MAE) metrics

Position (cm)
yxyy
0.340.17
DOI: https://doi.org/10.14313/jamris-2026-004 | Journal eISSN: 2080-2145 | Journal ISSN: 1897-8649
Journal RSS Feed
Language: English
Page range: 41 - 52
Submitted on: May 18, 2024
|
Accepted on: Nov 7, 2024
|
Published on: Mar 31, 2026
Published by: Łukasiewicz Research Network – Industrial Research Institute for Automation and Measurements PIAP
In partnership with: Paradigm Publishing Services
Publication frequency: 4 issues per year
Keywords:
Personalized rehabilitation,
Learning from demonstrations,
Dynamic Movement Primitives,
Model adaptive control
Related subjects:
Computer sciences,
Artificial intelligence,
Engineering,
Electrical engineering,
Control engineering, metrology and testing,
Mechanical engineering,
Fundamentals of mechanical engineering

© 2026 Ismail Auta, Ahmed Fares, Hiroyasu Iwata, Haitham El-Hussieny, published by Łukasiewicz Research Network – Industrial Research Institute for Automation and Measurements PIAP
This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License.

Previous articleVolume 20 (2026): Issue 1 (March 2026)Next article