Design and Modeling of an Upper Extremity

Exoskeleton
Salam Moubarak, Minh Tu Pham, Thomas Pajdla, Tanneguy Redarce

To cite this version:

Salam Moubarak, Minh Tu Pham, Thomas Pajdla, Tanneguy Redarce. Design and Modeling of an
Upper Extremity Exoskeleton. 11th International Congress of the IUPESM : medical physics and
biomedical engineering world congress 2009, Sep 2009, Munich, Germany. pp.476-479. �hal-00443902�

HAL Id: hal-00443902

https://hal.science/hal-00443902
Submitted on 5 Jan 2010

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est

archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents
entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non,
lished or not. The documents may come from émanant des établissements d’enseignement et de
teaching and research institutions in France or recherche français ou étrangers, des laboratoires
abroad, or from public or private research centers. publics ou privés.
 Design and Modeling of an Upper Extremity Exoskeleton
S. Moubarak1, M.T. Pham1, T. Pajdla2 and T. Redarce1
1
Laboratoire Ampere UMR CNRS 5005, INSA-Lyon, F-69621, France
2
Czech Technical University Prague, Department of Cybernetics, Center of Machine Perception, Czech Republic

Abstract— This paper presents the design and modeling simultaneously and the total cost of the therapy can be
results of an upper extremity exoskeleton mounted on a wheel reduced.
chair. This new device is dedicated to regular and efficient The paper is organized as follows: Section 2 deals with
rehabilitation training for weak and injured people without the human arm anatomy. Section 3 describes the different
the continuous presence of a therapist. The exoskeleton being a
wearable robotic device attached to the human arm, the user
components of our prototype. Section 4 presents the
provides information signals to the controller in order to kinematic and dynamic modeling of the exoskeleton.
generate the appropriate control signals for different training Finally, a brief conclusion highlights the ongoing works but
strategies and paradigms. This upper extremity exoskeleton also the perspectives of this project.
covers four basic degrees of freedom of the shoulder and the
elbow joints with three additional adaptability degrees of
freedom in order to match the arm anatomy of different users.

Keywords— Upper extremity exoskeleton, medical robotics,

rehabilitation.

I. INTRODUCTION

Elderly people, stroke victims, hemiplegic subjects,

paralyzed and spinal cord injured persons suffer from many
forms of handicaps for locomotion because of the decline of Fig. 1 The upper extremity exoskeleton
their physical and muscular strength. Robotic assistance and
rehabilitative training have recently become more common,
and thus, commercial robotic devices and orthosis are II. HUMAN ARM ANATOMY
already available and increasingly needed to assist, facilitate
and accelerate the training and the recovery of weak and The shoulder and elbow gather 5 bones of the upper
injured people. limb: the clavicle, the scapula, the humerus, the ulna and the
Several studies have been carried out on robotic medical radius [6] (Fig. 2).
training and assistance. For lower limbs, these works not
only concentrated on the therapeutic applications of the
robotic devices, but also on the development of lower
extremity exoskeletons for the augmentation of the load
carrying capability and the endurance of the user [1].
However, for the upper limbs, most researchers have paid
increasing attention to develop upper extremity
exoskeletons for medical assistance and rehabilitation
training [2]-[6].
This paper deals with the development and the modeling
of an upper extremity exoskeleton mounted on a wheel
chair (Fig. 1). Our objectives are to provide an efficient
rehabilitative training and physiotherapy for the patient’s Fig. 2 The shoulder and elbow bones
arm, as well as the possibility for the therapist to apply
different training strategies with a variable assistance degree From a biomechanical point of view, the human anatomy
depending on the state of the user and the injury level. can be represented as a set of rigid bodies connected by
Furthermore, this device enables the patient to have a joints. In robotic rehabilitation, biomechanical models can
consistent training without the continuous assistance of the be used for the design as well as the control of the prototype
therapist; therefore, the latter can work on many patients
Design and Modeling of an Upper Extremity Exoskeleton 477

to simplify the interaction mechanism between the human The 4 drives, the force sensors and the adaptability motors
and the robot. Mechanical analyses are carried out to are connected to a Dspace card and continuously
validate the kinematic structure of the robot and its exchanging information and control signals. Different
actuators dimensions. control schemes can be created and applied using Matlab /
Biomechanical models of the upper limb are also used Simulink and Dspace -Control Desk interface.
for evaluation and diagnosis, in order to control the forces
and the couples generated by the upper limb movements III.1- Mechanical structure:
during a robotic assisted rehabilitation session. The main purpose of an exoskeleton is not only to
provide efficient motion assistance to the human limbs, but
also to guarantee the safety and the comfort of the user.
That is why matching the human body anatomy is one of
the most important criteria for an exoskeleton design.

Fig. 3 The four basic degrees of freedom of the arm

The human arm has three complex articulations; the

shoulder, the elbow and the wrist. Our prototype, being
dedicated to the shoulder and elbow rehabilitation training,
covers three shoulder degrees of freedom (DoFs) (abduction
/ adduction, flexion / extension and internal / external
rotation) and one elbow DoF (flexion / extension) (Fig. 3).
Fig. 4 The mechanical structure of the exoskeleton
Table 1 Human arm characteristics
Shoulder Shoulder Shoulder Elbow The mechanical structure of our prototype (Fig. 4)
adduction flexion rotation flexion
mainly consists of three links and four revolute joints
Range of motion 0°– 180° -50° – 180° -80° – 100° 0° – 145° covering the basic degrees of freedom of the human arm.
The motion range of the robot’s joints is limited to provide
Table 1 gives the average range of the shoulder and the a wide risk-free workspace for the user. The arm is attached
elbow joints of the human arm [6]. to the exoskeleton by external arm and wrist holders with
pressure adjustable internal pneumatic holders. The rotation
motion is transmitted from the motor axes to the joint axes
III. A NEW UPPER EXTREMITY EXOSKELETON by rigid and compact spiro-conical gear systems. The
prototype is relatively lightweight (10 Kg) with a high ratio
This section describes the three prototype’s components: of DoFs / weight (Fig. 5).
-The mechanical structure
-The actuators and the force sensors
-The control system

The prototype has 4 motorized degrees of freedom; the

shoulder abduction / adduction, flexion / extension and
internal / external rotation as well as the elbow flexion /
extension motion. The shoulder height, the shoulder width
and the arm length can be adapted to fit with different body
sizes for different users. An arm holder and a wrist holder
with 8 integrated force sensors are sending information
signals to the controller in order to estimate the state of the
user and his motion intentions. Each motor is connected to a
drive and can be controlled either by torque or by velocity.
Fig. 5 The gear systems and the arm holders of the exoskeleton
478 S. Moubarak et al.

III.2- Actuators and force sensors: motors and the 8 force sensors are connected to the same
The four active DoFs of the prototype are motorized by electronic interface card.
brushless motors with high torque and relatively small The electronic card exchanges information and control
dimensions. The motor torques are amplified by the signals with a Dspace controller analyzing all the feedback
reducers and the gear systems, to give final output torques information and sending the appropriate control signals to
of 63 Nm for the shoulder abduction / adduction, 52.5 Nm the system depending on the training strategy. At this stage,
for the shoulder flexion / extension, 17.1 Nm for the different control schemes and therapy modes are created
shoulder internal / external rotation, and 13.5 Nm for the and applied to the system using Matlab / Simulink software
elbow flexion / extension motion. and Controldesk interface.
The prototype must be adaptable so it can be used by The complete electrical setup diagram of the system is
different patients with different body sizes. The shoulder shown in Fig. 7.
height, the shoulder width and the arm length can be varied
using three DC motors to fit with the users’ body
dimensions.
The control of the prototype with different training
strategies is mainly based on the information feedback
signals from the user and the exoskeleton. The motor
encoders’ feedbacks provide a real time estimation of the
position and the velocity of the joints and can be used as
feedback information for the kinematic and dynamic
models. However, to control an exoskeleton for medical and
rehabilitation applications, we need to estimate the state of
the user, his interactions with the robot and sometimes we
need to predict his motion intentions in order to apply the
appropriate robotic assistance. One solution consists in
using force sensors. Therefore, depending on the training
strategy, we can guide the patient’s arm in the correct
displacement or modify the exoskeleton’s motion to assist
him in his desired motion.
Fig. 7 The electrical setup

IV. MODELING OF THE EXOSKELETON

In order to achieve a compliant controller for the

exoskeleton, the kinematic and dynamic models of the
structure must be calculated and identification experiments
must be carried out to get a better estimation of the dynamic
and friction parameters of the robot, and therefore, improve
the theoretical models.

Fig. 6 The arm holder with the force sensors IV.1- Geometric and inertial parameters:
The mechanical structure of our prototype mainly
In our prototype, 4 force sensors are applied in each arm consists of three links and four joints. The geometric
holder. Each sensor can measure any force in the 0 – 125 N parameters of the robot have been calculated using the
interval. The sensors are surrounding the arm and the wrist Denavit-Hartenberg notations.
at 90° on from the other, detecting almost all the possible The theoretical inertial parameters, used in the dynamic
interactions between the user and the robot (Fig. 6). modeling have been calculated based on a complete CATIA
model of the prototype.
III.3- Control system:
All brushless motors are equipped with high resolution IV.1- Kinematic and dynamic models:
magnetic encoders for position feedback measurements. The following models have been calculated using
Each motor is connected to a drive with a sampling SYMORO+ software (Symbolic Modeling of Robots) and
frequency of 5 KHz. The 4 drives, the 3 adaptability DC simulated with Matlab/Simulink interface [7]:
Design and Modeling of an Upper Extremity Exoskeleton 479

- The direct kinematic model (DKM) Identification experiments are being carried out in order
- The inverse kinematic model (IKM) to improve the theoretical estimations of the inertial and the
- The direct differential kinematic model (DDKM) friction parameters of the exoskeleton.
- The inverse differential kinematic model (IDKM)
- The direct dynamic model (DDM)
- The inverse dynamic model (IDM) V. CONCLUSION AND FURTHER WORKS

The simulation diagram of the kinematic and dynamic This paper introduces the design and modeling results of
models is shown in fig. 8. an upper extremity exoskeleton mounted on a wheel chair.
A motion generator (MG) creates, for each joint, a 5th The key objective of this device is to help users with
degree displacement between two given positions, in a disabled upper limbs to practice their daily living activities
given time, starting and ending with a null acceleration. by providing different training levels and therapy modes. It
q , q and q represent the joint position, velocity and can accelerate the muscle strength development and the
recovery of post-stroke and spinal cord injured persons, as
acceleration vectors created by the motion generator. well as accident victims and elderly people.
X , X and * represent the final link position and The direct and inverse kinematic, differential kinematic
velocity vectors and the joints torque vector calculated by and dynamic models of our prototype have been created
the direct models. based on the Denavit-Hartenberg notations, calculated using
Q , Q and Q
 represent the joint position, velocity and SYMORO+ software, simulated and validated with
Matlab/Simulink interface.
acceleration vectors calculated by the inverse models.
Another aspect of this work concerns the control
strategy. The force sensors feedback will be used in order to
estimate the state and the motion intentions of the user; and
create efficient training strategies and therapy modes for a
q X Q better and faster recovery.
q
ACKNOWLEDGMENT
q This work was supported by EC project MEST-CT-2005-021024
WARTHE.
q X Q

REFERENCES
1. R. Steger, S.H. Kim, H. Kazerooni, “Control scheme and networked

q control Architecture for the Berkeley lower extremity exoskeleton
 (BLEEX)”, IEEE International conference on robotics and
* Q automation, Orlando, Florida, Mai 2006, pp.3469-3476.

q 2. E. Rocon, A.F. Ruiz, F. Brunetti, and J.L. Pons, “On the use of an
active wearable exoskeleton for tremor suppression via biomechanical
loading”, IEEE International conference on robotics and automation,
Orlando, Florida, Mai 2006, pp.3140-3145.
Fig. 8 The modeling and simulation diagram 3. K. Kiguchi, Y. Imada, and M. Lianage, “EMG-Based Neuro-Fuzzy
Control of a 4DOF power-assist exoskeleton”, 29th IEEE EMBS
International conference, Cité internationale, Lyon, France, August
The six kinematic and dynamic models have been 2007, pp.3040-3043.
4. R. Song, K.Y. Tong, X.L. Hu, S.F. Tsang, and L. Li, “The therapeutic
simulated and validated (e1 = e2 = e3 = 0). effects of myoelectrically controlled robotic system for persons after
An important criterion for comfortable rehabilitation stroke-A pilot study ”, 28th IEEE EMBS International conference,
training is the efficient gravity compensation of the New York city, USA, August-September 2006, pp.3504-3511.
prototype so that the user can move his arm freely without 5. M. Mihelj, T. Nef, and R. Reiner, “ARMin II-7DOF rehabilitation
robot: mechanics and kinematics”, IEEE International conference on
feeling the mass of the robot. The static component of the robotics and automation, Romania, Italy, April 2007, pp.4120-4125.
inverse dynamic model is applied on the prototype and 90% 6. A. Denève, “Développement et commande d'un robot pour la
of the gravity effect of the exoskeleton has been rééducation des membres supérieurs”, Ph.D. dissertation, University
compensated successfully. of Reims Champagne-Ardenne, France, 2007.
7. W. Khalil, E. Dombre, “Modélisation, identification et commande des
robots”, PARIS, Hermès, 1999.