Wheelchair Assistance with Servo Braking Control Considering Both

the Gravitation-Negating and the User’s Intention-based Assistance

Daisuke Chugo

, Nobuhiro Goto

, Satoshi Muramatsu

, Sho Yokota

and Hiroshi Hashimoto

Graduate School of Science and Technology, Kwansei Gakuin University, Sanda, Hyogo, Japan

School of Information Science and Technology, Tokai University, Hiratsuka, Kanagawa, Japan

Faculty of Science and Engineering, Toyo University, Kawagoe, Saitama, Japan

Advanced Institute of Industrial Technology, Shinagawa, Tokyo, Japan

Keywords: Manual Wheelchair Assistance, Passive Robotics, Servo Brake, Gravitation-Negating Control Algorithm,

User’s Intention-based Control Algorithm.

Abstract: This paper proposes a novel driving-assistance system for manual wheelchairs with consideration of both

uphill and downhill conditions. On an inclined road, there is a high risk of a wheelchair moving in a

direction that the user does not intend. In our previous works, the user has driven our assistive wheelchair in

the usual manner. Our proposed system estimates its user’s intentions and passively works to complement

their intentional force by negating the wheel traction that is generated by the road’s inclination using only

the servo brakes on each wheel. Nevertheless, in some cases, our system fails to assist the driving motion of

its user because the user drives the wheelchair in several ways that depend upon the environmental

condition, for example, during uphill or downhill driving. The required assistance is not constant according

to the situation, and it is difficult to assist with one wheel-control algorithm. Therefore, in this study, we

first investigate the required assistance condition according to the driving situation by conducting a

preliminary experiment with wheelchair users. Considering the results of this investigation, we then propose

a novel user interface that intuitively shows the system information and a wheel-control algorithm that

selects a suitable wheel controller according to the driving situation.

1 INTRODUCTION

Wheelchairs are widely used by mobility-impaired

people in their daily activities. In recent years, many

serious wheelchair-related accidents have been

reported. In Japan, more than 80% of wheelchair

accidents are caused by environmental hazards

(National Consumer Affairs Center of Japan, 2002).

The inclination of a sidewalk poses a potentially

high risk for a wheelchair user. The Japanese

government permits an incline in a sidewalk of up to

5°(Japan Institute of Construction Engineering,

2008). This inclination could potentially lead to a

wheelchair deviating from the sidewalk into the

roadway, which may result in collisions between

wheelchairs and cars. Therefore, a wheelchair

driving-assistance system is important for use on an

inclined sidewalk.

In previous research, many assistive technologies

for wheelchairs have been developed. Several

disabled people traditionally use power wheelchairs

(Yamaha Motor Co., Ltd., 2014) and previous

researchers have attempted to develop assistance

functions by adding wheels with actuators and

controlling them using robotic technology such as

motion control (Miller and Slack, 1995), sensing,

and artificial intelligence (Katevas et al., 1997)

(Murakami et al., 2001). These intelligent

wheelchairs provide several functions such as

suitable motion, obstacle avoidance, and navigation;

thus, they provide a maneuverable system. However,

many wheelchair users have the upper body strength

and dexterity to operate a manual wheelchair. For

these wheelchair users, such systems may be

excessively expensive and unnecessary.

Therefore, we have developed a passive driving-

assistance system for a manual wheelchair that uses

servo brakes (Chugo et al., 2015) (Chugo et al.,

2013). This system incorporates the concept of

passive robotics (Hirata et al, 2007). Our proposed

system passively operates on the basis of external

forces imposed by its user. No actuators are required

Chugo, D., Goto, N., Muramatsu, S., Yokota, S. and Hashimoto, H.

Wheelchair Assistance with Servo Braking Control Considering Both the Gravitation-Negating and the User’s Intention-based Assistance.

DOI: 10.5220/0005980503350343

In Proceedings of the 13th International Conference on Informatics in Control, Automation and Robotics (ICINCO 2016) - Volume 2, pages 335-343

ISBN: 978-989-758-198-4

335

in our system; it uses servo brakes, which can

control the brake torque, to produce the desired

motion according to the applied force and reference

track. In our previous research, we have developed

two wheel-control algorithms. One estimates the

intended direction of a manual wheelchair user by

determining the characteristics of hand motion and

maintains it as the reference track (Chugo et al.,

2013). The other simply negates the effect of

gravitational force on the wheelchair on an inclined

road (Chugo et al., 2015).

However, in some cases, these wheel controls

cannot assist in wheelchair driving, because users

row in different ways according to the

environmental situation, resulting in different

required assistance conditions. In particular, when

going uphill or downhill, a wheelchair driver uses

completely different driving techniques to control

their wheelchair. Under these conditions, our

controller cannot use only one wheel-control

algorithm. Therefore, in this study, we first

investigate how users drive their wheelchairs

according to the environmental situation and what

conditions are required for assisting these

techniques. Second, using the results of this

investigation, we propose a novel human interface

based on a hand brake and a wheel-control scheme

that combines a gravitational negating control

algorithm and a user’s intention-based control

algorithm. Using this idea, our proposed wheelchair

can continuously assist users in driving on uphill or

downhill roads.

This paper is organized as follows. We introduce

our assistive wheelchair and its problem

specification in section 2. In section 3, we propose a

novel human interface for our system and in section

4, we propose an improved driving-assistance

scheme based on the environmental situation. We

show the results of experiments using our prototype

in section 5. Section 6 presents our conclusions.

2 PROBLEM SPECIFICATION

ON OUR SYSTEM

2.1 System Configuration

Figure 1(a) shows our prototype wheelchair, which

utilizes a type of servo brake known as a powder

brake. Powder brakes are widely used in industrial

applications and their cost is low compared with

other servo brakes. The powder brake (Fig. 1(b))

(ZKG-YN50, Mitsubishi Electric Corp.) generates

enough brake torque to stop a wheelchair moving at

4 km/h, and containing a 100 kg user within 1 s. Our

prototype is based on a normal manual wheelchair

(BM22-42SB, Kawamura Cycle Co. Ltd.) and

fulfills the ISO7193, 7176/5 standards. Furthermore,

our prototype utilizes an encoder in each wheel to

measure the wheel-rotation velocity and two tilt

sensors in its body to measure roll and pitch angle

(see Fig.3).

(a) Overview (b) Installed Servo Brake

Figure 1: Our Prototype.

2.2 Problem Specifications in Daily

Usage with Our Assistive

Wheelchair Prototype

2.2.1 Preliminary Experimental Setup

Eight subjects attempted to navigate the test course

(Fig. 2) with our assistive wheelchair in order to

investigate how users drive their wheelchairs

according to different environmental situations. The

length of this course is about 1.5 km. The

experimental field is on the Kobe-Sanda Campus,

Kwansei Gakuin University, Japan. Our campus is

located atop a hill and this test course has uphill and

downhill roads to easily investigate problems with

our assistive system.

In this preliminary experiment, our system offers

assistance using two wheel-control algorithms; one

is a gravitation-negating control algorithm (Chugo et

al., 2015) and the other is a user’s intention-based

control algorithm (Chugo et al., 2013). In both cases,

our wheelchair system records all logs measured by

the equipped sensors and outputs brake traction

information. Furthermore, we record the subject’s

motion with a video camera during this experiment.

This preliminary experiment includes eight subjects

(Table 1), six wheelchair users and two able-bodied

people including a nursing specialist and a student.

Each subject tries one round using each algorithm.

ICINCO 2016 - 13th International Conference on Informatics in Control, Automation and Robotics

336

Flat

Uphill

Downhill

Flat (In school

cafeteria)

Start and

Goal position

100m

Figure 2: A test course for the preliminary experiment.

Table 1: Subjects.

Subject Age Sex

Weight

(kg)

Dominant

hand

Wheelchair

User*

A 66 Male 54 Right Yes

B 72 Male 62 Right Yes

C 68 Male 73 Left Yes

D 67 Female 51 Right Yes

E 74 Female 49 Left Yes

F 35 Male 81 Right Yes

G 39 Male 75 Right No

H 21 Male 59 Right No

* A subject who uses a manual wheelchair in daily life.

2.2.2 Problems at the User Interface

Table 2 shows major problems in this preliminary

experiment. Problem nos. 1, 2, and 3, which occur in

both control algorithms, can be solved by an update

of the user interface. For example, as problem no. 1,

the subject pushes the start button of our assistive

wheelchair system and tries to row its hand rim.

However, as the subject moves their hand from the

button to the rim, the wheelchair moves under

gravitational force on an inclined road. Furthermore,

as problem no. 2, our system stops the wheelchair

for safety reasons and an LED alerts the user that the

emergency brake is now working. However, on

many occasions, users cannot recognize this alert

and try to continue to drive the wheelchair.

Questionnaire results show that many subjects feel

that the system information is indistinct.

2.2.3 Problems with the Gravitation-

Negating Control Algorithm

Problem no. 4 is caused by the gravitation-negating

Table 2: Results of the Preliminary Experiment.

Contr

ol* Major Problems (Times)

Subjects

A B C D E F G H

1 I

The wheelchair moves by the

gravitational force after the user

pushes the start button. 1 2 1 1 2 1 1 0

I The user tries to move the

wheelchair when our system

stops by emergency brakes.

3 4 3 2 6 2 2 0

G 2 3 3 2 4 3 1 1

G The wheelchair moves by the

gravitational force after the user

switch off our system.

1 1 1 1 1 1 1 1

I 1 1 1 1 1 1 1 1

4 G

The user feels the wheelchair is

too heavy on an uphill situation.

(In some cases, the user cannot

go by own physical strength.) 6 7 7 10 14 4 6 3

5 I

The user cannot go the intended

direction on a downhill

situation. 8 7 8 9 11 6 7 1

6a I

Our system misjudges its user’s

intention. 9 5 8 8 11 8 8 4

6b I

Our system misapplies

emergency brakes. 3 3 2 3 3 2 2 0

* G is a gravitation-negating control algorithm and I is a user’s

intention-based control algorithm.

wheel-control algorithm. This wheel-control

algorithm negates the effect of the gravitational

force on the wheelchair on an inclined road. When

the user goes uphill on a road as in Fig. 3, the

wheelchair moves to a lower direction because of

the gravitational force on the inclined road. In this

condition, without an assistance system, a manual

wheelchair user should row the left wheel hard as

ff >

in Fig. 3(a) (where

is the row force at the

right wheel,

is the row force at left wheel.)

, f

The wheelchair tends to

move this direction

because of the gravity.

(a) Side view (b) Top view without assistance (c) Top view with assistance

Figure 3: Brake tractions on an inclined road.

To negate this gravitational force, our wheelchair

controls the servo brake according to (1) and (2),

where

(

)

yx ,

is the position of the center of gravity,

m is mass of the wheelchair and T is the width

between the wheels. Details regarding this calculation

were given in our previous paper (Chugo et al., 2015).

Wheelchair Assistance with Servo Braking Control Considering Both the Gravitation-Negating and the User’s Intention-based Assistance

337

()

elsef

ffif

mgy

clcr

sin

−

(1)

()

elsef

ffif

mgy

clcr

sin

−

−=

(2)

In the case of Fig. 3(b), our system generates the

brake traction,

, on a left wheel to negate the

gravitational force that leads the wheelchair to a

lower direction (a right direction). In this case, our

wheelchair user should row each wheel equally as

grgl

ff =

(where

is the row force at the right

wheel with our assistance and

is the row force at

the left wheel with our assistance). This means that

the user can row the wheelchair as if on a flat road;

however, a passive system does not assist the force

and the required row force increases with the brake

force,

, on the left wheel. Therefore, the users

feel as if the wheelchair is too heavy in an uphill

situation during this preliminary experiment.

2.2.4 Problems with the User’s

Intention-based Control Algorithm

Problem nos. 5, 6a, and 6b are caused by the user’s

intention-based algorithm. This algorithm uses

knowledge of neurophysiology in the form of the

minimum jerk trajectory model (Seki and Tadakura,

2004), which expresses the characteristics of hand

motion. According to this model, hand motion is

defined by equations (3) and (4), where x is the

position of the wheelchair,

is the initial position,

and

is the time when the user starts to row the

hand rim.

()

(

)

(

)

()

133

1683

10156

234

345

+−+−−+

+−−+=

mmm

mmmf

ttttt

tttttx

tttxxxtx





(3)

()

tttt

−=

(4)

In this model, unknown values include the end

position,

, and the final time,

. This algorithm

estimates two values by determining the

characteristics of hand motion when the user starts

to row a hand rim (at 0.1 sec). Figure 4(a) shows the

wheel velocity and the estimated movement using

this model. According to this method, our system

estimates the velocity of each wheel (

: velocity of

the right wheel,

: velocity of the left wheel) and

evaluates the intended direction of a manual

wheelchair user as Fig. 4(b). After estimation, the

system maintains the reference track that is

estimated until its user rows the wheel again. Details

regarding this algorithm were given in our previous

paper (Chugo et al., 2013).

However, in problem no. 5, our system fails to

assist when its user changes its movement direction

on a downhill road. When a user goes down an

inclined road, they turn by grasping a hand rim as a

brake, rather than by rowing the hand rim. The

control algorithm estimates the intended direction of

the user only when they accelerate the wheelchair by

their hand motion. When the user tries to change the

running direction by grasping a hand rim, our

assistive wheelchair controls the brake traction for

maintaining the reference track when the user

accelerates. Therefore, the assistance brake traction

by this wheel controller interferes with its user’s

intention.

Problem nos. 6a and 6b are parameter-setting

problems concerning how much error our system

accepts at the estimation of a user’s row motion. If

our system does not accept a larger error, its wheel-

control accuracy will increase; however,

misjudgement will also increase because of a

wheelchair vibration due to the unevenness of a road

surface. Thus, this is a trade-off problem.

0.1

0.2

0.3

0.4

0.5

0.6

0123

Time (sec)

Velocity (m/sec)

Expe riment

Minimum Jerk Model

Estimation

Time (0.1s)

Estimation

velocity fits

the real

velocity

profile.

Wheelchair

Human

(a) Minimum jerk model (b) Kinematics of our wheelchair

Figure 4: A user’s intention-based control algorithm.

3 PROPOSED USER INTERFACE

Based on the results of a questionnaire administered

to the subjects and the opinions of the nursing

specialist, the user interface of the assistive

wheelchair should have the following conditions:

 The input device should be equipped around a

hand rim, because the user activates the

assistive device and then rows a hand rim.

ICINCO 2016 - 13th International Conference on Informatics in Control, Automation and Robotics

338

Therefore, the distance between the input

device and the hand rim should be small.

 Subjects require very little information, namely

(1) whether a driving-assistive system works or

does not work, and (2) whether an emergency

brake works. Thus, its user interface should

clearly show this information.

Therefore, we propose a novel user interface

based on a hand brake as shown in Fig. 5.

Hook operated

by a Solenoid

Brake Pad

Driving

Assistance

is ON.

Driving Assistance is

OFF or an emergency

brake works.

Figure 5: A proposed user interface based on a hand brake.

Usually, a wheelchair user takes off a hand brake

when they drive, and then puts it back on when they

stop. Therefore, our system can determine the

intention of its user by the position of the hand brake.

The proposed user interface is quite simple; when

the user takes off a proposed hand brake interface,

our system starts offering driving assistance, and

when they put on a hand brake, our system stops

offering assistance. Furthermore, when our system

uses an emergency brake, this hand brake interface

moves to the off position automatically so that its

user can know easily that the emergency brake is

working.

The proposed hand brake uses a spring as in Fig.

5. A spring connected to a hand brake pulls it into

the off position. When a hand brake is in this

position, it pushes the brake pad to the wheel with

()

14≈

, and this force is the same as that of a

typical hand brake on a general manual wheelchair.

When the user switches our system on, they turn the

hand brake interface to the on position. There is a

hook with a rotational spring (Fig. 6) at this position

that holds this hand brake in place. The force

required to turn our system off is

()

8.0<

which is a light load for a manual wheelchair user.

When our system uses an emergency brake, a

solenoid equipped on a hook works as shown in Fig.

6(b) and releases the hand brake. The hand brake is

backed to the off position by the spring. Figure 7

shows our prototype hand brake interface, which

moves to the switch off position automatically.

The proposed hand brake interface works as a

normal hand brake, meaning the user can simply

replace an original hand brake on a general

wheelchair with the proposed hand brake interface.

This mechanism fulfills the ISO7193, 7176/5

standards and can be installed on a general

wheelchair without any special reconstruction.

Solenoid OFF

Solenoid ON

Rotational

Spring

Brake Lever

(a) The solenoid is off.

(b) The solenoid is on.

Figure 6: A hook operated by a solenoid.

(a) The hand brake is off.

(b) The hand brake is on.

Figure 7: A prototype of a proposed hand brake interface.

4 PROPOSED WHEEL CONTROL

ALGORITHMS

4.1 Combination of the

Gravitation-Negating Control

Algorithm and the User’s

Intention-based Control Algorithm

From the results of a preliminary experiment, the

major driving techniques and required assistance

conditions for the wheelchair user are as follows.

 The wheelchair driving technique consists of

two phases—a rowing phase and an inertial

running phase.

 In the rowing phase, the load should be small,

especially in an uphill situation. Based on the

opinions of the wheelchair users in the

preliminary experiment, no brake traction is

felt to be better than gravitational cancellation.

Wheelchair Assistance with Servo Braking Control Considering Both the Gravitation-Negating and the User’s Intention-based Assistance

339

The nursing specialist thinks that when the

wheelchair goes in an uphill direction, its user

concentrates on rowing its hand rim and

cancels the gravitational force unconsciously

due to inclination. The gravitational

cancellation makes users spend their physical

strength on the brake traction.

 In the rowing phase on a downhill situation, the

gravitational force should be removed for safe

driving. Based on the opinions of wheelchair

users, on a downhill road, the required force to

row is small and the wheelchair tends to

deviate from the intended direction of its user

due to gravitational force, and users report

fearing this motion.

 In the inertial running phase, the wheelchair

deviates from the intended track due to the

gravitational force; therefore, driving

assistance is necessary. However, in many

cases, wheelchair users grasp the hand rim and

change the running direction.

Therefore, we propose a novel wheel-control

scheme that combines the gravitation-negating

control algorithm and the user’s intention-based

control algorithm as follows.

 For reducing the required physical strength in

an uphill situation, our system uses the user’s

intention-based control algorithm during the

rowing phase.

 For the same reason, on a flat floor situation,

our system uses the user’s intention-based

control algorithm during the rowing phase.

 To increase the driving ability in other

situations, our system uses a gravitation-

negating control algorithm during the rowing

and inertial phases on a downhill road and the

inertial phases on uphill and flat roads.

 For safety reasons, when the wheelchair

accelerates in all situations, our system judges

whether this acceleration is done by human

rowing motion. If not, our system turns on an

emergency brake.

 When our system switches to a different

control algorithm, it controls the brake traction,

ref

, according to (5) to prevent sudden

change:

()

⎪

⎩

⎪

⎨

⎧

→+

−

→+

−

GIif

IGif

ref

(5)

where

is the brake-traction reference derived by a

gravitation-negating control algorithm (Chugo et al.,

2015) and

is a reference by the user’s intention-

based control algorithm (Chugo et al., 2013).

the switching time between the two control

algorithms, which we set to 0.1 s in this study.

IG →

means that our system switches from a

gravitation-negating control algorithm to a user’s

intention-based control algorithm.

Figure 8 shows the details of the proposed

algorithm. Our system measures the road inclination,

, using a tilt sensor and evaluates the uphill or a

downhill condition.

START

Does the wheelchair

accelerate?

Does the acceleration fit

the characteristics oh a

human movement?

Does the road incline?

A gravitation negating

control algorithm

A user's intention based

control algorithm

Emergency

Brake

A downhill

An uphill or a flat

Does the wheelchair re-

accelerate?

Does the wheelchair re-

accelerate?

Yes Yes

Yes

No Yes

Figure 8: Flow chart of our proposed control scheme.

4.2 Parameter Setting for Estimation of

a User’s Rowing Motion

Our system judges that wheelchair acceleration is

caused by human rowing motion if the difference

between the real velocity and human movement

profiles is less than the pattern-matching parameter,

, in the user’s intention-based control algorithm

(Chugo et al., 2013). However, the parameter should

change according to the road condition.

In the preliminary experiment, the wheelchair

accelerates 3,700 times; 3,043 of these are caused by

the user’s rowing motion. Our system evaluates these

accelerations with various parameters and the

evaluation results are presented in Fig. 9. The false

positive error is the misjudgement of human motion

as acceleration by some other source and the false

negative error is the misjudgement of the acceleration

by other sources as being due to human motion. From

the results, our system can distinguish between being

ICINCO 2016 - 13th International Conference on Informatics in Control, Automation and Robotics

340

indoors or indoors based on unevenness in a road

surface, and uses

0.3

in an outdoor environment

and

0.2

in an indoor environment.

20%

40%

60%

80%

100%

0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

Threshold (C

)

Rate [%]

Negative Failure

Positive Failure

Success

20%

40%

60%

80%

100%

0.511.522.533.544.55

Threshold (C

)

Rate [%]

Negative Failure

Positive Failure

Success

(a) Asphalt surface (outdoor) (b) Linoleum floor (indoor)

Figure 9: Success rate with each pattern-matching

parameter.

5 EXPERIMENTS

5.1 Experimental Setup

We tested our system’s performance in two

experiments. In the first experiment, the subjects

move from side to side in a figure of eight on a test

road with an 8° incline using our prototype

wheelchair with the proposed controller (Case P) as

in Fig. 10. In this course, (I), (III), and (V) in Fig.

10(a) are uphill and (II) and (IV) are downhill. To

verify the controller’s effectiveness, the subjects

repeated this activity in wheelchairs without the

system (Case N), with only a gravitation-negating

control (Case G) and with only the user’s intention-

based control (Case I). The subjects are the same as

those of the preliminary experiment as shown in

Table 1. In the second experiment, subjects try the

test course shown in Fig. 2 with our proposed

system. All experimental conditions are the same as

in the preliminary experiment.

8deg

Start and Goal

Position

(I)

(III)

(II)

(IV)(V)

(a) Test course (b) Real environment.

Figure 10: Test course on an inclined road.

5.2 Experimental Results

The results show that the subjects could drive in an

intended direction when using our system (Fig. 11).

Figure 12 shows the running tracks of the

wheelchair. With the proposed assistance system,

the subject can drive the wheelchair smoothly. On

the other hand, in case I, it is difficult to change the

forward direction on a downhill situation and in case

G, the brake traction that negates the gravitational

force increases the load in uphill situations and it is

difficult for the subject to climb in the vertical

direction. As a result, in case G, the subject makes a

detour.

(a) Passing (II) (b) Passing (III)

Figure 11: Test run with our proposed controller by

subject H.

-2.5

-2

-1.5

-1

-0.5

0.5

1.5

-10123456789

[m]

Proposed Scheme

O nly Inte ntion

Only Gravity

Without Assistance

The subject cannot change its

direction at Case I.

A hand rim is too heavy and

the subject cannot climb to the

vertical direction at Case G.

The wheelchair deviates to the

intended direction by the

gravity force at Case N.

Moving direction

Pole

Figure 12: Running tracks by subject H.

Figure 13 shows the brake traction differences

between the right and left wheels. A positive value

means that a brake traction on the right wheel is

generated and our system negates a gravitational

force to the left direction. A negative value implies

the opposite. In Fig. 13, for example, when a

wheelchair passes (I), our system negates a

gravitational force to the right direction; thus, the

traction value is negative. On the other hand, the

Wheelchair Assistance with Servo Braking Control Considering Both the Gravitation-Negating and the User’s Intention-based Assistance

341

traction value is zero when the subject rows a hand

rim because at this time, our system uses the user’s

intention-based control algorithm. Furthermore,

Table 3 shows the workload that a subject outputs

during one trial. Our proposed scheme requires only

the workload of the user’s intention-based control

algorithm. From these results, our proposed control

scheme realizes a gravitation-negating function with

a smaller workload.

Table 4 shows the experimental results with our

proposed user interface and the proposed controller

on the test course shown in Fig. 2. The proposed

user interface works effectively and settles major

problem nos. 1, 2, and 3. The subjects can use the

proposed interface without difficulties. The

proposed control scheme settles major problem nos.

4 and 5. Subjects D and E are women with

-100

-50

100

0 5 10 15 20 25

Time [sec]

Torque [Nm]

III

The system switches to an

intention based control and stops

to negate the gravity force.

-100

-50

100

0 5 10 15 20 25

Time [sec]

Torque [Nm]

III

The system switches to an

intention based control and stops

to negate the gravity force.

Figure 13: Brake traction differences in case P by subject

Table 3: Workload for one trial (J).

Subject

A B C D E G H

Proposed Scheme 457.1 445.6 498.4 403.2 418.5 453.7 447.0

Only Intention 442.6 428.6 471.1 389.2 398.2 424.6 426.1

Only Gravity 523.4 493.5 552.1 427.5 466.2 479.1 501.4

Table 4: Results on a test course with the proposed system.

No Major Problem (Times)

Subjects

A B C D E F G H

The wheelchair moves by the

gravitational force after the user pushes

the start button. 0 0 0 0 0 0 0 0

The user tries to move the wheelchair

when our system stops by emergency

brakes. 0 0 0 0 0 0 0 0

The wheelchair moves by the

gravitational force after the user switch

off our system. 0 0 0 0 0 0 0 0

The user feels the wheelchair is too

heavy on an uphill situation. 0 0 0 3 2 0 0 0

The user cannot go the intended direction

on a downhill situation. 0 0 0 0 0 0 0 0

6a Our system misjudges its user's intention. 1 0 0 1 0 0 0 0

Our system misapplies emergency

brakes. 2 1 1 2 2 0 1 1

somewhat less physical strength in their hands who

feel that an uphill road is a heavy load using a

normal manual wheelchair. Although the accuracy

of rowing motion-estimation increases, there are

some errors due to small steps.

From these results, we can verify that our

proposed system is effective for assisting a manual

wheelchair.

6 CONCLUSIONS

This paper presents an investigation into the way

wheelchair drivers operate in various environmental

conditions. We explore several assistance strategies

that are appropriate to the various operating modes

that the wheelchair driver presents with in these

conditions. By this investigation, we propose a novel

human interface based on a hand brake and a wheel-

control scheme that combines a gravitational

negating control algorithm and a user’s intention-

based control algorithm. Using this idea, our

proposed wheelchair can continuously assist users in

driving on uphill or downhill roads. Its effectiveness

in daily usage is verified by experimental results

with our prototype.

ACKNOWLEDGEMENTS

This work is supported in part by Kawanishi

Memorial ShinMaywa Education Foundation and

Exploratory Research on Feasibility Study (FS)

Stage (AS242Z00295K) by Adaptable and Seamless

Technology Transfer Program through Target-driven

R&D, Japan Science and Technology Agency (JST).

REFERENCES

National Consumer Affairs Center of Japan, 2002.

Considering the safety of self-propelled manual

wheelchairs, In NCAC News, Vol.14, No.3, pp.4.

Japan Institute of Construction Engineering, 2008. A

Sidewalk guideline for smooth mobility, 2nd edition,

ISBN: 4802893922. (in Japanese).

Yamaha Motor Co., Ltd., 2014. JW Swing,

http://global.yamaha-motor.com/ymgn/group_topics/2014/11_03/

D. P. Miller and M. G. Slack, 1995. Design and Testing of

a Low-Cost Robotic Wheelchair Prototype, In

Autonomous Robots, Vol.2, pp.77-88.

N. I. Katevas, et al., 1997. The Autonomous Mobile Robot

SENARIO: A Sensor-Aided Intelligent Navigation

ICINCO 2016 - 13th International Conference on Informatics in Control, Automation and Robotics

342

System for Powered Wheelchairs, In IEEE Robotics

and Automation Magazine, Vol.4, No.4, pp.60-70.

Y. Murakami, Y. Kuno, N. Shimada and Y. Shirai, 2001.

Collision avoidance by observing pedestrian’s faces

for intelligent wheelchairs, In Proc. of Int. Conf. on

Intelligent Robots and Systems, pp.2018-2023.

D. Chugo, N. Goto, S. Muramatsu, S. Yokota, and H.

Hashimoto, 2015. Robotic Driving Assistance System

for Manual Wheelchair User on Uneven Ground, Proc.

of 2015 IEEE Int. Conf. Rehab. Robot, pp.648–653.

D. Chugo, T. Higuchi, Y. Sakaida, S. Yokota, and H.

Hashimoto, 2013. A Driving Assistance System for a

Manual Wheelchair using Servo Brakes, Proc. of 10th

Int. Conf. on Informatics in Control, Automation and

Robotics, pp.259-266.

Y. Hirata, A. Hara and K. Kosuge, 2007. Motion Control

of Passive Intelligent Walker Using Servo Brakes, In

IEEE Trans. on Robotics, Vol.23, No.5, pp.981-990.

H. Seki and S. Tadakura, 2004. Minimum Jerk Control of

Power Assisting Robot based on Human Arm

Behavior Characteristics, In Proc. of IEEE Int. Conf.

on Systems, Man and Cybernetics, pp.722-727.

Wheelchair Assistance with Servo Braking Control Considering Both the Gravitation-Negating and the User’s Intention-based Assistance

343