Mechanical Analysis for Active Movement of Upper Limb Rehabilitation Robots to Alleviate Shoulder Pain in Patients with Stroke Hemiplegia and Frozen Shoulder

TL;DR

This paper presents a rehabilitation robot designed to help stroke patients with shoulder pain by replicating natural shoulder movement for effective upper-limb recovery.

Contribution

A novel rehabilitation robot with biomechanically accurate kinematic design to alleviate shoulder pain in stroke patients.

Findings

The robot's forward kinematic analysis showed average positional errors between 0.5% and 2.8%.

Inverse kinematic analysis confirmed stable convergence to target positions using the Gauss–Newton method.

The robot accurately reproduces shoulder motion within the designed range of motion.

Abstract

Shoulder disorders, including frozen shoulder resulting from stroke-induced hemiplegia, significantly reduce a patient’s ability to perform activities of daily living, thereby necessitating repeated rehabilitation. Consequently, extensive research has been conducted on rehabilitation robots to assist in upper-limb motor recovery. The shoulder moves according to the scapulohumeral rhythm. Considering the biomechanical characteristics of the shoulder joint, the rehabilitation robot was designed to replicate a similar kinematic environment using actuators and linkages that emulate the structures of the upper arm, shoulder, and clavicle. To ensure precise operation, the kinematic accuracy of the robot was pre-evaluated. Kinematic analyses were conducted using MATLAB, and the results were compared with coordinate data from the mechanical design to evaluate positional accuracy. In addition, the convergence and accuracy of joint-angle estimation for target positions were analyzed. The forward kinematic analysis revealed that the average positional error between the measured and target coordinates ranged from 0.5% to 2.8%, with the Base Motor–Back Motor segment exhibiting the highest error (2.8%). The inverse kinematic analysis demonstrated stable convergence to the target positions through iterative computations using the Gauss–Newton method, confirming that the actual motion could be accurately reproduced within the designed range of motion.

Figures