Design and Control of a Force Reflecting Arm Exoskeleton for

Virtual Reality Applications

Dimitar Chakarov, Ivanka Veneva, Mihail Tsveov and Dimitar Trifonov

Institute of Mechanics, BAS, Acad. G. Bonchev str. Block 4, Sofia, Bulgaria

Keywords: Exoskeleton Arm, Design, Haptic Device, Impedance Control, Pneumatic Muscle, Joint Torque, Antagonistic

Interaction, Virtual Gymnastics.

Abstract: In this paper, the design of an exoskeleton for the upper limb is presented, aimed primarily at training and

rehabilitation in virtual environments. A mechanical model of the exoskeleton arm as haptic device is built

up and impedance control scheme is selected as the most suitable for force reflection at the arm. The design

of a grounded exoskeleton prototype is revealed in the paper. A driving system based on braided pneumatic

muscle is selected to ensure natural security in the interaction. Antagonistic drive system for each joint is

shown, using pulley and Bowden cable transmissions. An approach is presented for the joint moments control

by antagonistic interaction of bundles with different numbers of pneumatic muscles. Control scheme of joint

torque by antagonistic interaction is given, too. Computer simulations are performed to provide power

reflection by virtual reality (VR), according to scenario of virtual gymnastics.

1 INTRODUCTION

Exoskeleton is outside wearable robot with joints and

limbs corresponding to those in the human body.

Active exoskeletons transmit their driving

mechanisms human joints moments. Exoskeletons

are used in four main functions in accordance with

their control algorithms (Perry, 2007): a)

Rehabilitation - related closely to the body, they

perform tasks of physical therapy in active or passive

operation; b) Haptic device - people physically

interact with virtual objects, as the forces of

interaction apply to humans through the exoskeletons

actuators; c) Master device - at control type "master-

slave", the forces of interaction apply to exoskeleton

- "master"by the robot-"slave"; d) Assisted human

device - amplifier of the human body, the operator

feels lighter loads, adopted by exoskeleton.

With the increase in recent years of applications,

related to interaction in virtual environments

increases the importance of the second function:

using exoskeleton as Haptic device. Haptic interfaces

allow humans to perceive and transmit forces and

movement of real and virtual entities and actually to

be "immersed" in virtual reality. It is possible to give

people the illusion, that they have another body in

immersion in virtual reality. When you move you see

the virtual body to move in the same way (Slater,

2014), through motion capture in real time. This

function for using exoskeletons is combined

effectively with their other application - as a device

for rehabilitation, using virtual scenarios.

In literature are presented many exoskeletons

possessing very different mechanical structure and

drive, with or without force feedback effect. The first

modern exoskeleton of human upper limb was

designed by PERCRO to feel and response in contact

and collision with real and virtual entities

(Bergamasco, 1994). The authors of PERCRO

developed researches with designing other

exoskeleton of human upper limb for Haptic

interaction with virtual environments L-Exos (Frisoli,

2009).

Exoskeleton is known with pneumatic actuators,

developed by (Lee, 1998). Their device is with 9

degrees of freedom and allows full playback of

human hand workspace of exoskeleton. Alternative

exoskeleton is developed by (Jeong, 2001), which is

designed to overcome the limited bearing capacity of

previous projects, using parallel mechanisms and

pneumatic drives. Other examples of exoskeletons,

targeted primarily for rehabilitation are ARMIN of

the University of Zurich (Nef, 2007) and LDCs at the

University of Salford, which uses pneumatic

actuators (Tsagarakis, 2003). Comfortable to wear

Chakarov, D., Veneva, I., Tsveov, M. and Trifonov, D.

Design and Control of a Force Reﬂecting Arm Exoskeleton for Virtual Reality Applications.

DOI: 10.5220/0006427403350342

In Proceedings of the 14th International Conference on Informatics in Control, Automation and Robotics (ICINCO 2017) - Volume 2, pages 335-342

ISBN: Not Available

335

hand of Salford overcomes some of the problems and

limitations of previous designs. This soft hand-

exoskeleton is used for physiotherapy and for

training.

These devices have to meet the safety

requirements except the traditional requirements for

performance. It is important to develop exoskeletons

possessing naturally low impedance in order to

achieve natural safety in the mutual interaction “man-

robot”.

One of the most common approaches to

implement natural compliance is the usage of

pneumatic artificial muscles (Daerden, 2002),

(Tsagarakis, 2003), (Caldwell, 2007). The pneumatic

muscles are actuators with a high power/weight ratio

and can be directly coupled to the joint, without a

heavy and a complex gearing mechanism. The force-

deflection characteristic of these muscles, being

strongly non-linear, they can be used to achieve

compliance adjustment by means of an antagonistic

setup. The drawbacks of the pneumatic muscles are:

slow response to the input control, the presence of

hysteresis, making precise position control difficult to

realize and bringing the necessity of pressurised air.

Although various solutions are developed in order

to be achieved simultaneous safety and performance

within the human connected devices, there are still

significant barriers that should be overcome (such as

structure, drive and sense) to be able to provide

natural, safe and comfortable physical interaction on

the one hand and performance on the other. The

objective of this work is to introduce the design of an

active arm exoskeleton, ensuring natural, safe and

comfortable physical interaction with human and to

display force-related sensory information from a

virtual environment to the user. The exoskeleton is

designed primarily for training and rehabilitation in

virtual environments.

2 MODELING AND CONTROL

OF EXOSKELETON ARM

The device, which can be seen to function as a

powered exoskeleton, is build-up of a branched serial

structure, kinematics similar to the structure of the

human body as it is shown in Figure 1.

The kinematics structure consists of four movable

bodies 1, 2, 3, 4 and four rotational joints (Figure 1).

The first two rotational joints J1, J2 are incident and

mutually orthogonal in order to emulate the

kinematics of a universal joint with the same centre

of rotation as the human shoulder. The third and the

fourth joints J3 , J4 are selected to be two rotational

joints incident and mutually orthogonal in order to

emulate the kinematics of a universal joint with the

same centre of rotation as the human elbow flexion

and human shoulder rotation. Moreover, the third

joint was assumed to be coincident with the elbow

joint as it is shown in Figure 1. The effective

movements of so selected four joints simulate the

movements of human arm in the shoulder and elbow.

The exoskeleton upper limb structure has h = 4

degrees of mobility. This structure is chosen in order

to construct an arm exoskeleton consisting of two

equal type actuated universal joints.

J1

J2

J3

J4

EE

2

1

3

4

0

Figure 1: Structure scheme of exoskeleton arm.

Mechanical model of the upper limb exoskeleton

is build up, according to the kinematics scheme

shown in Figure 1. Generalized parameters are

accepted to be the parameters of relative movements

in the joints of exoskeleton represented by the (h x 1)

vector q. End-effector (EE) of exoskeleton arm

moves in the ν - dimensional operation space.

Assuming that the operator hand is connected at a

point to the exoskeleton end effector, further we will

consider exoskeleton operation space as 3 -

dimensional. The coordinates of the exoskeleton end-

effector are presented by (3 x 1) vector X. The relation

between the generalized parameters q and the

coordinates of the end-effector X is set by the direct

kinematic problem. Regarding velocities, we

formulate that problem by means of equality:

qJX 

(1)

Above J is the (3 x 4) matrix of Jacoby.

To make the operator feel the simulated

dynamics, impedance control is selected as the most

suitable for force refection between the user and the

virtual environment. The impedance model in

Cartesian space is given by the following equation of

simplified dynamics

FX

Z







(2)

Above

ICINCO 2017 - 14th International Conference on Informatics in Control, Automation and Robotics

336

T

]XX;X;X[X

0



(3)

is vector of end effector motion parameters, and

]K;B;M[Z 

(4)

is the mechanical impedance of the system in

Cartesian space including the matrix of inertia,

damping and stiffness M, B, and K.

An open-loop impedance controller with model-

feedforward is selected (Carignan, 2000). This type

of force control is called open-loop because there is

no force feedback from the device to the controller to

regulate the force output of the exoskeleton end

effector as it is illustrated by the control block

diagram in Figure 2. The goal of the model

feedforward, is to cancel out the corresponding terms

in the dynamics of the device. In control block

diagram of Figure 2 the feedforward terms for the

device stiffness and gravity are included. This scheme

allows the use of so called strategy “patient-in-

charge” when the patient lead the robot with a low

impedance K

d

and resistive forces are generated for

deviations from the target in the virtual scene.

+

∆Q

Q

d

J

T

G

+

Q=J

T

F

-

+

Δ

X

d

-

Δ

X

Exoskeleton.

joint control

F

d

Human

f

e

K

Δ

q

X

c

d

K

J

1

e

K



Q

f

VR

scenario

Δ

X

3D Unity

Virtual Engine

Figure 2: Block scheme of open-loop impedance controller

with feedforward stiffness compensation.

According to this diagram, the controller takes the

difference between the desired position deflection

∆X

d

=X

d

−Xo, and the actual position defection of the

haptic interface, ∆X ≡ X – Xo. Since the positions on

this linearized block diagram are represented by

deviations from some nominal position Xo, the

Cartesian position of the virtual interface and the joint

position of the exoskeleton ∆q are related through the

Jacobian J. In this diagram, the exoskeleton natural

stiffness is represented by joint-space stiffness K

e

which is the relationship between the joint torque

inputs and the differential angular output. The

following relationships can be obtained from the

block diagram

QGQQQ

f

d



(5)

]QGQQ[JKX

f

de



1

(6)

Above

]XX[KJFJQ

d

c

d

T

d

T

d



(7)

represents desired force command in joint space.

Similarly, in joint space the forces that the human

exerts on the exoskeleton end effector are generated

as:

FJQ

T



(8)

Note, that Q represents a physical, not control input

to the device. The term Q

f

in (5) represents force

command in joint space as a result of model

feedforward compensation. It is calculated in

accordance with modelled exoskeleton stiffness in

joint space K

e

f

according to equation

XJKqKQ

f

e

f

e

f



1

(9)

The (h x 1) vector of (5):





T

G

h1

G,...,G

(10)

is the vector of gravity torques, generated at the

exoskeleton joints. Its components:











n

1i

j

i

T

oij

q

gmG

, j = 1 ,…, h (11)

are determined by the mass of links m

i

, i=1…n, the

gravity acceleration vector in the base frame

g

0

=[g

x

,g

y

,g

z

]

T

and the position vector of the i-th link

centre of mass in the base frame ρ

i

= [ρ

x

, ρ

y

, ρ

z

]

T

. After

substituting equation (7), (8) (9) and (10) into (6) and

setting ΔX

d

=0, without loss of generality yields:

FGJX]KKK[

Tcf

e

c

d

c

e





(12)

Above

1



 JKJK

e

T

c

e

represents the natural

stiffness of the exoskeleton in Cartesian space and

1



 JKJK

f

e

T

cf

e

represents the stiffness model of

the exoskeleton in Cartesian space. The influence of

the member

]KK[

cf

e

c

d



, that represents the active

exoskeleton stiffness decreases, if the natural stiffness

of the exoskeleton

c

e

K

is close to the desired level

c

d

c

e

KK 

. The aim is to develop an exoskeletons

possessing naturally low impedance suitable for

interaction with virtual objects where the value of the

desired environment stiffness is low.

Design and Control of a Force Reﬂecting Arm Exoskeleton for Virtual Reality Applications

337

3 MECHANICAL DESIGN AND

ACTUATION SYSTEM OF

EXOSKELETON ARM

The mechanical structure of exoskeleton system

includes two arms, that are grounded, as each arm is

built up, according to structural diagram if Figure 1.

Each arm has four active degrees of freedom,

corresponding to the natural motion of the human arm

from the shoulder to the elbow. Structure in Figure 1

is chosen in order to develop a powered exoskeleton

arm, in which two equal type universal joints with 2

DoF are used and so, unlike other solutions, circular

guide and three axes joints are avoided.

Each universal joint placed in the shoulder and in

the elbow consists of a cross shaft and pair of

rotational joint oriented at 90 ° to each other. The

cross shaft is formed as an angle, in which the human

hand is situated, as the axes of rotation intersect at a

point, coinciding with the point of rotation of the

human natural joint.

The basic parts of the exoskeleton arm structure

are made of aluminium. The arm is constructed for

use by a typical adult. All arm units are designed that

allow quick and easy adjustment of length, so making

it easy to accommodate a range of users. Plastic shells

are placed over the units with bands for attachment to

human limb. This design suggests a comfortable, light

and low cost mechanical structure. In Figure 3, the

structure of the right arm exoskeleton is shown, as a)

is a computer simulation, and b) a picture of arm

prototype. Exoskeleton left hand is designed as a

mirror image of the right hand and is mounted on the

common fixed base.

a) b)

a) computer simulation, b) picture of arm prototype.

Figure 3: Structure of the right exoskeleton hand:

Masses of the exoskeleton arm units are defined using

CAD program. The masses of four basic units (Figure

1) are as follows m

1

=0.437 kg, m

2

=0.594 kg,

m

3

=0.364kg, m

4

=0.434kg. Lengths of exoskeleton

arm and forearm in the initial setup are as follows

0.286 m and 0.370 m respectively. The exoskeleton is

designed so, that it covers the requirements of

“Activities of daily living” (ADLs) as they have been

assessed in (Perry 2007). The ADLs range of motions

(ROM) and exoskeleton joint ranges of motions are

shown in Table 1.

Actuation system of exoskeleton arm should have

the following advantages: excellent power/weight

ratio with inherent safety, natural compliance, low

cost. To achieve these advantages self-made braided

pneumatic muscle actuators (PMA) are used. PMA

consists of two layers: an inner one, representing

rubber liner and outer one, representing braided

nylon. Endcaps are allocated at both ends, to which

are clamped by clips the two layers. In one endcap is

located a pipeline for supplying the pressurized air.

Muscles possess a maximum diameter D

0

= 0.016 m

and nominal length L

n

=0.390 m. Thus created PMA

are used not only singly, also in a sheaf of several

pneumatic muscle, as shown in Figure 4.

Muscles from the bundle are fed in parallel with

air at a maximum pressure 600 kPa (6 bar). Supply

pipeline is connected with two parallel arranged

valves of the type on / off, one of which serves to

supply the maximum pressure and the other for

discharge to the atmosphere. Pressure sensors are

connected to the supply pipeline at every muscle

bundle. All the muscles of the bundle are connected

mechanically in one and at the other end, where they

attach themselves.

Figure 4: Sheaf pneumatic muscle actuators.

Joint motion/torque on the exoskeleton arm is

achieved by antagonistic actions through cables and

pulleys, driven by the pneumatic actuators. Two

sheafs PMA, a and b, work together in an antagonistic

scheme, simulating a biceps-triceps system to provide

the bidirectional motion/force, (Figure 5).

0

a

Ln

L

C

max

L

min

b

P

b

S

P

а

T

1

T

2

q

r

Figure 5: Antagonistic scheme for joint motion/torque.

ICINCO 2017 - 14th International Conference on Informatics in Control, Automation and Robotics

338

Cable transmissions T

1

and T

2

are used for the

coupling between the muscles and the pulley. The

pulleys are fastened on the arm segments following

joint shafts. All actuators are mounted on the exo-

shell on the operator’s back. High precision rotation

sensors S are mounted in the joints.

According to the indications in Figure 5, muscle

contraction is defined as:

C = L

n

– L (13)

Muscle contraction C is typically presented as a

percentage of the nominal muscle length L

n

determined by the outer braided layer c=100 C/L

n

.

Figure 6: Graphs force / contraction of bundles of 1…6

muscles and their replicas at supply pressure of 400 kPa.

Experiments were conducted under static load of the

used PAM, as the number of muscle in a bundle and

the magnitude of the supply pressure has been altered.

At set pressure, the contraction of muscles is

measured by their nominal value L

n

to a minimum

value L

min

.

Experimentally obtained force/ contraction

diagrams (1m , … , 6m) for the used bundles

consisting of 1 to 6 muscles at constant supply

pressure of 400 kPa, are shown in Figure 6. At zero

pressure the muscles show elastic properties

determined of inner rubber liner. Pneumatic muscle

behaves like a pressure dependent variable

compliance spring. Graphs force/contraction of

braided pneumatic muscle actuators typically

represent quadratic functions. For replication of the

used PAM as a nonlinear quadratic spring (Caldwell,

2007) can be used the following equation

)LL(LKP

minp



. (14)

where: P is the pooling force of the muscles, L and

L

min

are the muscle lengths according Figure 5, K

p

is

a pressure dependent coefficient. Taking into account

(13) and equality for maximum contraction of muscle

bundle C

max

= L

n

– L

min

, equation (14) can be

transformed

)CC)(CL)(tpk(P

maxno



(15)

where: p is the operating pressure, k

o

and t are empiric

derived coefficients that depend on the muscle

number.

Approximated characteristics are composed for

bundles from different number muscles m = 1,2,3,4,5

and 6 at a pressure p = 100, 200, 300 and 400 kPa.

The maximum contraction of each muscle bundle is

c

max

= 40% Ln or C

max

=0.156 m. In the process of

approximation have been determined the following

relations of the coefficients k

o

and t in accordance

with the number of bundle muscles m: k

0

= -39 + 321

m, t = -0.21+ 4.41 m.

Approximated characteristics of muscle bundles

with 1, 2, ..., 6 muscles at a supply pressure of 400

kPa, are shown on Figure 6, (1m-a , … , 6m-a), where

they are compared with experimentally obtained

characteristics.

When the muscle bundles move the joint in an

antagonistic scheme (Figure 5), the current

contractions of each bundle in the joint C

a

, C

b

are

determined by the joint end positions q

max

, q

min

.

To generate the maximum forces, muscles are

attached so, as to operate with a minimum

contraction, i.e., in the end joint position, one bundle

has zero contraction and the contraction of the other

is determined according to the equation

*

q

rqC 

(16)

where r is pulley radius and q* = (q

max

- q

min

) is joint

range of motion. Contraction (16) can be estimated as

effective contraction of both muscles and to be

presented as a percentage of the nominal muscle

length L

n

:

n

*

q

L/rq100c 

. In each position of the

joint, the equation (16) represents the sum of the

preliminary contraction of the two muscle bundles.

In a current position q of the joint, the

contractions of the two muscle bundles are

)qq(rC

max

a



(17)

)qq(rCCC

min

aqb



(18)

The forces P

a

, P

b

of joint muscle bundles can be

calculated according to (15), (17) and (18). It is

assumed, that one muscle sheaf is active in one

direction, when the other is passive and vice versa at

the sign change of the torque, establishing a

maximum torque in the joint of antagonists muscle

actuators. The muscle bundles with zero pressure

always participate with a force in the joint

antagonistic equilibrium, as they are elastic.

Antagonistic balance in each joint is achieved by

0

200

400

600

800

0 5 10 15 20 25 30 35

P[N]

c%

6m

5m

4m

3m

2m

1m

6m‐a

5m‐a

4m‐a

3m‐a

2m‐a

1m‐a

Design and Control of a Force Reﬂecting Arm Exoskeleton for Virtual Reality Applications

339

generating desired torques in each joint, according to

equality:

)rP(PQ

abi



(19)

Since the bundle muscles with zero pressure (for

example a) p

a

= 0 generates elastic force P

a

according

to (15) end (17), draws (19), (15) and (18) allow to

calculate the required pressure p

b

of other muscles

(for example b)

)q,Q(fp

ib



(20)

According to control block diagram on Figure 2, the

joints torques Q

i

are set from equality (5) as

components of the vector representing force

command in joint space



T

4i1

Q,..,Q,...,QΔQ 

(21)

The control scheme of the joint torque caused by the

antagonistic interaction is shown in Figure 7.

Pressures in the muscle bundles p

a

and p

b

are

monitored by pressure sensors mounted to each

muscle sheaf.

PAM“a”

PAM“b”

Air

p

0

Presure

senso

r

“a”

Joint

Presure

sensor“b”

Position

sensor

p

b

Vending

Valve

Filling

Valve

Filling

Valve

Vending

Valve

Air

p

0

q

Q

i

p=0

-

+

p=0

p - p

a

-

+

-

+

p-p

b

∆Q

p

a



Figure 7: Control scheme of the joint torques.

A selection of bundles with different number of

muscles in every joint is performed to ensure nominal

torques in the joints and to provide joint moves q *

(Chakarov, 2014). Selected bundles antagonists with

number muscles m of the four arm joints are shown

in Table 1. The table shows also achieved torques for

each joint and muscles effective contractions.

For example, in Figure 8 is illustrated the change

of the moment in the shoulder joint, under the

influence of muscles antagonists a and b according to

contraction in abduction/adduction: a) in abduction is

active bundle b, which changes the pressure; b) in

adduction is active bundle a and its pressure is

changed. Shaded area of the chart shows the muscles

work area, corresponding to the stroke of the joint. In

this area, the achieved joint torque in according to

(19) depends on contraction and is amended for

shoulder abduction within the boundaries of 38.5 –

19.2 Nm, and for shoulder adduction within the

boundaries of 4.7 - 13.9 Nm. Similarly, graphics are

derived, amending moments in other joints. Table 1

shows the achieved torques for each joint.

a)

b)

Figure 8: Amendment of the torque in the shoulder joint

under the influence of muscle bundles antagonists a and b:

a) in abduction; b) in adduction.

4 COMPUTER EXPERIMENTS

FOR ENSURING FORCE

REFLECTION FROM VR

The exoskeleton is designed as a haptics device for

applications in virtual reality. For example, one

application is a virtual gymnastics. 3D Unity Virtual

Engine is used for data transfer between exoskeleton

and virtual entity (avatar) in virtual gymnasium.

Optical tracker HMD Oculus Rift is used for

visualization in the virtual system and to track the

head position and orientation. Exoskeleton is

designed to track position and orientation of human

body and to provide force feedback to human arm.

Figure 9 shows a scenario of virtual gymnastics - an

exercise with fitness cable machine.

When the user set optical tracker attached to his

head and "puts on the exoskeleton" as attaches the

upper limbs to both arms of exoskeleton, he is

"immersed" in the virtual scene gymnastics. Man gets

the ability to transmit movement to the avatar and

takes strength from it. One sees the virtual body to

move like him, through motion capture in real time.

0

10

20

30

40

‐20 0 20 40

Q[Nm]

C%

ma=3,mb=7

pa‐0

pb‐0

pb‐

200[kPa]

pb‐

600[kPa]

pb‐

400[kPa]"

0

10

20

30

40

‐25 ‐5 15 35

Q[Nm]

C%

ma=3,mb=7

pa‐0

pa‐

200[kPa]

pa‐

400[kPa]

pa‐

600[Kpa]

pb‐0

ICINCO 2017 - 14th International Conference on Informatics in Control, Automation and Robotics

340

Table 1: ADLs and exoskeleton range of motions, effective muscle contractions, number of muscles and achieved torques.

Bundles

antago

nists

ADLs range

of

motions

[deg]

Exoskeleton

range of

motions

[deg]

Effective

muscle

contrac

tions

c%

Number of

the

muscles

in the

bundle

Achieved

torques

[Nm]

Shoulder

J1

Abduction b 100

110 15.5 7 38.5 – 19.2

Adduction a 15.5 3 4.7 - 13.9

Shoulder

J2

Flexion b 110

120 16.9 6 32.9 – 16.2

Extension a 16.9 3 5.2 - 14.8

Elbow

J3

Flexion b 150 150 21.1 4 21.9 – 7.8

Extension a 21.1 2 2.0 – 9.8

Elbow

J4

Int. rotation b 135 135 19.0 4 21.7 – 7.2

Lat. rotation a 19.0 3 4.3 – 15.8

The controller in Figure 2 creates power

commands in the joints of exoskeleton, that form the

force action on the user's hands, equivalent to the

power impact simulated in the virtual scene. In the

scene of virtual gymnastics with cable machine,

power influence is the tensile strength, which cable

exerts on hand. The magnitude of this force is a

constant, determined by the weight of the virtual

weight stacks. For this scene, virtual machine

generates a desired force command F

d

in the

controller in Figure 2, without taking into account the

desired Cartesian stiffness K

d

.

Figure 9: Scene of virtual gymnastics - an exercise with

cable machine.

Simulations were conducted for assessing the

projected exoskeleton ability to provide forceful

impact on the operator, as the described scene of

virtual gymnastics. Cable machine and the man’s

hand are simulated, as shown in the way in Figure 10.

The figure shows the right arm in several positions,

and the cable attached to the EE and slung over a roll

in a fixed point T. Experiments were carried out in the

following manner. A tensile force F is set, which

virtual cable exerts on the avatar hand. This is the

force, that exoskeleton must exercise over the hand of

man. Exoskeleton arm performs a movement of

maximum tension within the boundaries of joint

ranges. Torques in the exoskeleton joints (5) were

calculated by taking into account the desired moment

d

T

d

FJQ 

, as well as joint component of gravity

compensation G. The desired force is assumed to be

equal to the cable force with reverse sign F

d

= -F.

Gravitational component (10) were calculated after a

preliminary masses and geometry assessment of the

exoskeleton movable units.

Figure 10: Simulation of the scene of virtual gymnastics

with cable machine.

All joints are driven by antagonist bundles,

consisting of different number muscles, performing

effective muscle contractions, as shown in Table 1.

Supply pressure for the active muscles changes on the

range of 0 – 600 kPa, until the pressure of the passive

ones is 0. The experiment shows that for values of the

tensile force F = 0-50 N, desired torques in the joints

of the exoskeleton are achievable within the limits of

the attainable torques, shown in Table 1. In Figure 11

are shown the results of the shoulder joint at the

movement "tension" with a force F = 50 N. It is shown

the change of the joint components of the moments

ΔQ, Q

d

, G, as well as the achieved maximum moment

in the joint Q

max

, which is a result of the antagonist

interaction of the bundles of muscles with maximum

Design and Control of a Force Reﬂecting Arm Exoskeleton for Virtual Reality Applications

341

and minimum pressure. On the X axis is given the

contraction of active muscles bundle.

Figure 11: Amendment the joint components of ΔQ, Q

d

, G,

Q

max

, in the shoulder joint at cable force of 50 N.

5 CONCLUSIONS

In this paper, the design of an active arm exoskeleton,

aimed at human interaction with virtual environment

is presented. In the work, a mechanical model of the

exoskeleton arm, as haptic device is built up and

impedance control scheme is selected for force

refection at the arm. The design of a grounded

exoskeleton prototype with low mass/inertial

characteristics is revealed in the paper. To ensure

natural security in the interaction, a driving system

based on braided pneumatic muscle is selected.

Antagonistic drive system for each joint is shown,

using pulley and Bowden cable transmissions.

An approach for joint moments control by

antagonistic interaction of bundles with different

numbers of pneumatic muscles is presented.

Computer experiments have been carried out to

provide force reflection by VR, according to scenario

of virtual gymnastics. The conducted simulations

show the possibility of developed exoskeleton to

provide impact strength of 50 N on the hand of the

operator. Experiments with real exoskeleton

prototype were conducted with volunteers, where

they evaluate their feelings and their embodiment in

a virtual scene. Additional measurements and

evaluations of physical quantities are going to be

performed.

ACKNOWLEDGEMENTS

This work was funded by the European Commission

through FP7 Integrated Project VERE - No. FP7 -

257695 and by Bulgarian Science Found, Call: 2016,

through Project AWERON – DN 07/9, to which the

authors would like to express their deepest gratitude.

REFERENCES

Perry J., Rosen J, Burns S., 2007. Upper-limb powered

exoskeleton design. IEEE/ASME Transactions on

Mechatronics. Vol.12, No4, August 2007, pp. 408 –

417.

Slater M., Sanchez-Vives, M.V., 2014. Transcending the

Self in Immersive Virtual Reality. IEEE Computer 47

(7): 24-30 (2014), DOI: 10.1109/MC.2014.198.

Bergamasco M., B. Allotta, L. Bosio, L. Ferretti, G. Perrini,

G. M. Prisco, F. Salsedo, And G. Sartini, 1994. An Arm

Exoskeleton System for Teleoperation and Virtual

Environment Applications, IEEE Int’l Conf. Robot.

Automat., vol. 2, 1449–1454.

Frisoli A., Fabio Salsedo, Massimo Bergamasco, Bruno

Rossi And Maria C. Carboncini, 2009. A force-

feedback exoskeleton for upper-limb rehabilitation in

virtual reality, Applied Bionics and Biomechanics, Vol.

6, No. 2, June 2009, 115–126.

Lee,S., S. Park, M. Kim, And C.-W. Lee, 1998. Design of

a Force Reflecting Master Arm and Master Hand using

Pneumatic Actuators. IEEE Conference on Robotics

and Automation, May 1998, 2574–2579.

Jeong Y., Y. Lee, K. Kim, Y. Hong, And J. Park, 2001. A

7 DOF Wearable Robotic Arm using Pneumatic

Actuators,” in Proc. of the 32

nd

ISR International

Symposium on Robotics, April 2001, 388-393.

Nef T., M. Mihelji, and R. Riener, 2007. Armin: a robot for

patient-cooperative arm therapy. Medical & Biological

Engineering & Computing, 45 :887–900.

Tsagarakis N.G., Caldwell Dg., 2003. Development and

control of a “soft-actuated” exoskeleton for use in

physiotherapy and training. Autonomous Robots.

15(1):21–33.

Daerden Fr., D. Lefeber, 2002. Pneumatic Artificial

Muscles: actuators for robotics and automation.

European Journal of Mechanical and Environmental

Engineering, 47(1), 1 - 11.

Caldwell D.G. et al., 2007. “Soft” exoskeletons for upper

and lower body rehabilitation — design, control and

testing. International Journal of Humanoid Robotics

Vol. 4, No. 3, 549–573.

Carignan, C.R., K. R. Cleary, 2000. Closed- Loop Force

Control for Haptic Simulation of Virtual Environments,

Haptics-e, Vol. 1, No. 2, February 23, pp.1-14.

Chakarov, D., I. Veneva, M. Tsveov, T. Tiankov, D.

Trifonov, 2014. New Exoskeleton Arm Concept Design

and Actuation for Haptic Interaction with Virtual

Objects, Journal of Theoretical and Applied

Mechanics, Sofia, vol. 44, No 4, pp. 3-14, DOI:

10.2478/jtam-2014-00.

0

5

10

15

20

25

30

35

40

024681012

[Nm]

c%

Qmax

Q

Qd

G

ICINCO 2017 - 14th International Conference on Informatics in Control, Automation and Robotics

342