6th Joint Seminar on Rehabilitation Engineering and Assistive Technology

6th Joint Seminar on Rehabilitation Engineering and Assistive Technology

Neural Electronic Engineering Laboratory
Division of Biomedical Engineering for Health and Welfare
Graduate School of Biomedical Engineering
Tohoku University
Sendai, Japan

Biocybernetics Laboratory
Department of Biomedical Engineering
Institut Teknologi Sepuluh Nopember (ITS)
Surabaya, Indonesia

Schedule:
Date: June 21, 2023
Time: 13.00 Surabaya /15.00 Sendai
Venue: Online (Zoom Meeting)

Presentation list:
1) Laily Asna Sagira, Achmad Arifin, Hendra Kusuma, Andra Risciawan, Moh.
Ismarintan Sazuli
Sliding Mode Approach to Design Adaptive Control Scheme for an Exoskeleton Robot Rehabilitation for Upper Limb Movement Restoration
2) Yudai Ohta, Takashi Watanabe
A study on a wrist joint model for FES Control test:  measurement of joint stiffness and introduction of short range stiffness
3) Nur Azizah, Achmad Arifin, Fauzan Arrofiqi A Preliminary Development of Hybrid FES-Robot Rehabilitation System for Restoring Upper Limb Movements
4) Shuta Kano, Takashi Watanabe
A Preliminary Study on Estimation of Voluntary Effort Using IMU Signals and Machine Learning for FES Rehabilitation

Sliding Mode Approach to Design Adaptive Control Scheme for an Exoskeleton Robot Rehabilitation for Upper Limb Movement Restoration

Laily Asna Safira¹, Achmad Arifin^1,2, Hendra Kusuma¹, Andra Risciawan³, Moh. Ismarintan Zazuli^1,3

¹Graduate Program of Electrical Engineering, Electrical Engineering Department, ITS, Surabaya, Indonesia

²Biomedical Engineering Department, ITS, Surabaya, Indonesia

³Manufaktur Robot Indonesia (MRI), Surabaya, Indonesia

Abstract

Spasticity of the upper limbs generally happens in stroke survivors, resulting in motor skills loss due to motor nerve problems. The rehabilitation process, with the proper intervention, can help to improve motor function. Exoskeleton robot is one of the upper limb rehabilitation methods that use external power to improve the effectiveness of the rehabilitation process in the repetitive movements on subjects. In order to use robotics as a rehabilitation tool safely and accomplish objectives, position control must be taken into consideration. The dynamics of patient movement during the rehabilitation process affect the movement of the DC motor as an actuator on the exoskeleton. The primary purpose of this paper is to enhance the robotic joint’s position control performance by applying adaptive PID control with online parameter tuning. The adaptation process used position error and sliding surface method. Implementation of this adaptive PID based on sliding surface verified by experiment involving six normal subjects who mimic hemiplegia in elbow joint movements control. Results showed that adaptive control can be applied to the exoskeleton best adaptation constant 0.1 and m constant 1.0. Root Mean Squared Error was 0.67±0.0 degree with Maximum Absolute Error of 0.98±0.0 degree. The stability analysis showed that the control results followed local asymptotic stability.

A study on a wrist joint model for FES Control test: Measurement of Joint Stiffness and Introduction of Short Range Stiffness

Yudai Ohta, Takashi Watanabe

Graduate School of Biomedical Engineering, Tohoku University, Japan

Abstract

Functional Electrical Stimulation (FES) is a useful technique to restore functional movement of upper limbs paralyzed. Joint stiffness plays an important role in maintaining a posture against disturbance. Therefore, the final goal of this study is to develop an FES controller that simultaneously controls joint angle and joint stiffness. In this report, first, the musculoskeletal model of the wrist joint for FES control test developed in our previous study was tested in calculating joint stiffness. In the experiment, one degree-of-freedom of wrist joint movement (dorsiflexion/palmar flexion) was measured by stimulating the flexor carpi ulnaris (FCU) and extensor carpi radialis longus/brevis (ECRL/B) with a healthy subject, in which a disturbance was applied to calculate the stiffness. The values of joint stiffness calculated from measured data were compared to those of model simulation. The stiffness values obtained from model simulation were differed significantly from the values measured with the subject. Although stiffness calculation method used in the model simulation was based on the slope of the length-force curve, recently the validity of a stiffness calculation method using Short Range Stiffness (SRS) has been reported. Therefore, a stiffness calculation method using SRS was introduced. As a result, the proposed method using SRS was suggested to show characteristics of the stiffness closer to the measured values than the conventional method.

A Preliminary Development of Hybrid FES-Robot Rehabilitation System for Restoring Upper Limb Movements

Nur Azizah, Achmad Arifin, Fauzan Arrofiqi

Biomedical Engineering Department, ITS, Surabaya, Indonesia

Abstract

Stroke is a condition caused by the blockage or rupture of an artery in the brain. One of the repercussions of stroke is extreme motor impairment of the upper extremities. To regain lost motor function, a rehabilitation process is essential. Common rehabilitation methods for post-stroke patients include the use of exoskeletons or Functional Electrical Stimulation (FES).  These methods can be synergistically combined into a hybrid system known as the Hybrid Exoskeleton-FES. In this study, we propose the preliminary development of a specialized Hybrid Exoskeleton-FES system dedicated to rehabilitating the elbow joint. The FES component of the hybrid system consists of a power supply, Boost Converter, and pulse generator. It provides subthreshold stimulation to deliver a therapeutic effect and enhance the effectiveness of motor rehabilitation. In this study, a closed-loop control system is implemented using a PID controller for the exoskeleton and a Fuzzy Logic Controller for the FES. The PID controller compensates for error values in motor movement, ensuring that the actual angle closely matches the desired angle along the target trajectory. On the other hand, the Fuzzy Logic Controller adjusts the amplitude of stimulation to produce movement amplitudes aligned with the target trajectory. Experiment tests involved five healthy subjects who performed elbow flexion and extension exercises with a sinusoidal trajectory (frequencies: 0.01 and 0.02 Hz) have been performed. Root Mean Squared Errors (RMSE) of hybrid operation with sub threshold electrical stimulation were 1.651±0.009 degree, and 0.087±0.01 degree, for 0.01 Hz and 0.02 Hz trajectory, respectively. These values were not significantly different from the RMSE of the exoskeleton only movement training. For the amplitude limit of electrical stimulation was 2 times the subthreshold of the Triceps and Biceps muscles, the RMSE was 1.650±0.003 degree. Results demonstrate the successful integration of the hybrid system, encompassing the exoskeleton and functional electrical stimulation, thereby enabling simultaneous operation.

A Preliminary Study on Estimation of Voluntary Effort Using IMU Signals and Machine Learning for FES Rehabilitation

Shuta Kano, Takashi Watanabe

Graduate School of Biomedical Engineering, Tohoku University, Japan

Abstract

Voluntary effort during rehabilitation using functional electrical stimulation (FES) would be effective in enhancing effectiveness of rehabilitation training in stroke patients. This study aimed to show the feasibility of estimating voluntary effort during FES-assisted movements using signals from inertial measurement units (IMUs). In this report, estimating the difference of effort was tested by classifying load condition under the same movement. First, elbow joint movement in the sagittal plane was measured under the 4 load conditions (no load, 2 kg, 3kg, and 5kg load) with a healthy subject, in which electrical stimulation with small amplitude was applied. The IMUs were worn on the upper arm and forearm of the subject’s left (non-dominant) arm. Then, classifiers were tested in the classification of the load conditions under different input signals. Three types of classifiers were compared: k-nearest neighbor, linear SVM and non-linear SVM, and four types of input signals were tested: raw signals (acceleration and angular velocity), frequency spectrum and time-frequency representation of a signal by short-time Fourier transform (STFT), or by continuous wavelet transform (CWT). The results showed that a linear support vector machine classifier with numerical values obtained from the time-frequency transformation of the raw signals by STFT or CWT as input achieved a classification accuracy exceeding 98% in classifying three load conditions.