An Observation of Behavioral Changes of Indoor Dogs in Response to

Caring Behavior by Humanoid Robots

Can Dogs and Robots Be Companions?

Motoko Suzuki, Yuichi Sei, Yasuyuki Tahara and Akihiko Ohsuga

Graduate School of Information Systems, The University of Electro-Communications, Chofu, Tokyo, Japan

Keywords: Pet, Animal, Robot, Interaction, Home.

Abstract: The aim of our research is to build good relationships between pets and robots at home. We aim to promote

of positive interaction between pets and robots. Recently, robots have been become popular with the general

populace. There is a lot of research in human-robot interaction. We pay attention to pets that live in houses

with humans. It is required for pets to like robots for positive interactions between pets and robots to exist. In

this paper, we examine that 1) a robot can take care of dog, and 2) dogs and robots can be companion by

caring behavior of robots toward dogs. In our experiment, we used two robots. One of the robots takes care

of a dog, while the other does not. We observed which robot the dog chooses to interact with and had seventeen

dogs participate in this study. We performed this statistical test to judge whether the dogs treated the robots

with any significant differences.

1 INTRODUCTION

Recently, communication robots like Pepper

(Aldenaran, 2015) have become popular in the

everyday household. Existing studies have been

exploring the possibility of coexistence between

humans and robots in a human-robot interaction field.

The coexistence of pets and robots that live in the

same, however, has not yet been explored. The

objective of this study is to disclose behavioral

guidelines for robots living with pets. We are

therefore researching Pet-Robot Interaction (PRIN)

(Figure 1). There are two main goals of PRIN: 1)

promotion of positive interaction and 2) prevention of

negative interaction. The positive interaction

represents the good relationships between pets and

robots. Examples of such positive interactions include

robots becoming pet playmates instead of humans

(Figure 2 left), robots feeding pets (Figure 2 right),

robots training pets, and robots aiding in the care of

the pets of the elderly. The negative interactions

represent the bad relationships of pets and robots.

Examples of such negative interactions include pets

that are surprised by the movement of robots and pets

disturbing the actions of robots. It is important to

disclose the factors that promote positive interactions

and prevent negative interactions for good

relationships between pets and robots to develop

through various situations at home.

Figure 1: Image of PRIN.

The study focused on two main research

questions:

RQ1 : Can robots take care of dogs?

RQ2 : Can dogs and robots be companion by caring

behavior of robots toward dogs?

In this study, we provided two conditions of

robots being around pets. The first was robots

performing caring behaviors toward pets, and the

second was robots performing no caring behaviors

toward pets involved in the study. As a result, caring

Suzuki M., Sei Y., Tahara Y. and Ohsuga A.

An Observation of Behavioral Changes of Indoor Dogs in Response to Caring Behavior by Humanoid Robots - Can Dogs and Robots Be Companions?.

DOI: 10.5220/0006188604810488

In Proceedings of the 9th International Conference on Agents and Artiﬁcial Intelligence (ICAART 2017), pages 481-488

ISBN: 978-989-758-220-2

481

behaviors of robots influenced the behavior of the

dogs.

In Section 2 of this paper, we describe the

implementation of movement of caring-behavior

robots and the preparations made for the experiment.

Section 3 discusses the procedures involved in the

experimental observation of behavioral changes in

dogs in response to caring-behavior robots. We

describe the results of the experiment in Section 4.

Section 5 breaks down a discussion of the experiment,

and next, Section 6 describes research related to our

experiment, such as Human-Robot interaction and

animal behavior. Lastly, Section 7 summarizes our

research with a conclusion.

Figure 2: The dog on the left is running after the ball thrown

by a robot. The dog on the right is eating food that was

given to it by robots.

2 THE METHODOLOGY OF

CARING BAHAVIOR OF

HUMANS AND HUMANOID

ROBOTS TOWARD DOGS

In Section 2, we explore the difference between when

a human take care of a dog and when the robot takes

care of a dog. In 2.1, we will observe the care of dogs

by humans, and in 2.2, we will describe parts of the

preparations undertaken for the experiment, including

the environment of the room, the implementation of

movement patterns of caring-bahavior robots, and the

result evaluation method.

2.1 Caring Behaviour of Humans

toward Dogs

Currently, dog owners are used to taking care of their

dogs all the time. Examples of cares given to dogs

include feeding, acting as a playmate, toilet cleaning,

taking for walks, brushing, bathing, clipping nails and

hair, etc. We think the burden can be reduced for the

owner by humanoid robots that can care for dogs. In

addition, we think that caring-behavior robot will

improve the dog’s quality of life. We will focus on

two aspects of the experiment: “playmate” (playing

ball) and “feeding”.

First, we observed the behaviour of each dog

when humans took care of it. Table 1 shows the

details of the participant dogs during this part of the

trial.

§1 Playmate (Playing Ball)

The experimenter threw the ball, and we observed

whether the dogs held the ball in their mouths. This

was repeated five times. As a result, nine dogs held

the ball in their mouths five times each, one dog held

the ball in its mouth one time, and seven dogs didn’t

hold the ball in their mouths at all. Through this part

of the trial, it became clear that there are dogs that

play ball and dogs that do not play ball.

§2 Feeding

Dogs B and E were unable to concentrate on this

part of the experiment. The experimenter fed using

tray the other dogs without trouble. We observed

whether the dogs ate. This was repeated five times.

As a result, fifteen dogs ate the human-introduced bait

five times each.

Table 1: Participant dogs.

ID Breed Sex Age

(in years)

A Shih Tzu F 9

B Pomeranian F 5

C Mix M 0.5

D Chihuahua M 4

E Mix F 0.5

F Mix F 0.5

G Mix F 0.5

H Toy Poodle M 4

I Mix M 3

J Mix F 0.5

K Chihuahua M 6

L Mix F 3

M Toy Poodle M 5

N Mix M 4

O Chihuahua F 6

P Toy Poodle F 3

Q Toy Poodle F 7

2.2 Caring Behaviour of Humanoid

Robots for Dogs

We observed the behaviour of the dogs when humans

took care of them in 2.1. Next, we will examine the

behaviour of the dogs when humanoid robots take

ICAART 2017 - 9th International Conference on Agents and Artiﬁcial Intelligence

482

care of them. Two robots were used in the

experiment. We defined one of the robots taking care

of a dogs as the “care robot”, and the other robot was

referred to as the “non-care robot”. In this

experiment, we observed behavioral changes in the

dogs in response to the caring behavior of the care

robot.

2.3 Preparation of Experimentation

In the experiment using the care-robot and non-care

robot, we will describe the implementation of

movement patterns of the caring-behavior robots in

2.3.1. The environment of the room is described in

2.3.2.

2.3.1 Implementation of Care Robot

This experiment involves the humanoid robot Pepper

(made by ALDEBARAN). Pepper’s movement

patterns were installed using Choregraphe 2.4.3

software. The two types of care behavior involved

("playmate" and "feeding") required the

implementation of human-like movement. The

following information is a detailed account of the

behavior of the robot.

§1 Playmate (Playing Ball)

The robot needed to throw the ball to play with

each dog. The motion of the robot was divided into

the following three actions for it to throw the ball.

STEP 1: The palm of the robot faced upward and

robot held the ball with its fingers bent, and the arm

bent at the elbow.

STEP 2: The robot swung the arm down while

holding the ball.

STEP 3: The robot threw the ball by opening the

fingers while swinging the arm up.

Figure 3 below shows the steps of the robot throwing

the ball.

Figure 3: The robot throws a ball.

§2 Feeding

We prepared a bait tray by tying a string to the

paper plate for the robot to the lift and lower the bait

(Figure 4). The robot fed the dog by placing the plate

on the floor using the string secured to the bait tray.

We use the bait that each dog eats normally.

Figure 4: Bait tray with a string to it.

The feeding motion of the robot was divided into

the following two actions.

STEP 1: The string of the bait tray was put in the palm

of the robot’s hand.

STEP 2: The robot put the bait tray on the floor by

bending at the waist.

Figure 5 below shows the steps of the robot feeding

the dog.

Figure 5: The robot puts the tray on the floor.

2.3.2 Experiment Room

The room was large enough to place the care robot

and non-care robot inside and a dog could move

freely. In the room, there were chairs and tables.

Figure 6 is a sketch of the room in which the

experiment took place. Diagonal lines show the

furniture. Smiling faces represent the position of the

robots. The two robots faced the center of the room.

Figure 6: Sketch of the room.

An Observation of Behavioral Changes of Indoor Dogs in Response to Caring Behavior by Humanoid Robots - Can Dogs and Robots Be

Companions?

483

3 EXPERIMENT

The experiment was to explore the non-care robot and

care robot actions implemented in Section 2 and

observe the dog’s behavioural changes following the

care robot’s actions toward the dogs. After the care

robot performed a care action, the tray containing the

bait was placed in front of the two robots. One dogs

at a time was released from a distance of 2 m. from

the robot, and we observed whether the dogs ate the

food in front of either robot.

In order from the dog to determine the difference

between the care robot and non-care robot, the dogs

distinguished between the yellow and blue (Neitz,

1989). One of the two robots was dressed in the

yellow outfit, and the other wore blue clothes. The

color of clothing was replaced for each test dog.

The experimenter changed the position of the

robot that took over the caring care behavior for each

seventeen dogs. The following describes the details

of the procedure for each caring behavior.

3.1 Playmate (Paying Ball)

Experiments were conducted in the following

conditions and procedures for all seventeen

participant dogs.

3.1.1 Conditions

The playmate phase involved the two conditions

below.

Condition 1: The care robot played ball with the

dog.

Condition 2: The non-care robot didn’t play ball

with the dog.

3.1.2 Procedures

First, the care-robot, non-care robot, and a dog were

placed into the experiment room (2.2.2). The

experiment followed four steps for each participant

dog (Figure 7).

STEP 1: Care robot threw the ball. Then we

observed whether the participant dogs held the ball in

its mouth. This was repeated 10 times.

STEP 2: Put the trays of bait in front of the two

robots.

STEP 3: We observed which tray the dog ate from

first. This was repeated 5 times.

STEP 4: The experimenter swapped the position

of the care robot and non-care robot manually in front

of the eyes of a dog.

STEP 5: The Tray of bait were put in front of the two

robots. We then observed which tray the dog ate from

first. This was repeated 5 times.

Figure 7: The procedures of the experiment.

ICAART 2017 - 9th International Conference on Agents and Artiﬁcial Intelligence

484

3.2 Feeding

Experiments were conducted in the following

conditions and procedures for fifteen of the

participant dogs. Exclude Dogs B and E because they

couldn’t concentrate on this part of the experiment.

Experiments were conducted following the

conditions and procedures described below.

3.2.1 Conditions

Two conditions of the feeding act are described

below.

Condition 1: The care robot feed the dog.

Condition 2: The non- care robot didn’t feed the

dog.

3.2.2 Procedures

Care robot, non-care robot, and a dog were placed the

experiment room (2.2.2). The experiment followed

the four steps below for each participant dog.

STEP 1: Care-robot fed the dog. This was

repeated 10 times.

STEP 2: Put the trays of bait in front of the two

robots.

STEP 3: We observed which tray the dog ate from

first. This was repeated 5 times.

STEP 4: The experimenter swapped the position

of the care robot and non-care robot manually in front

of the eyes of a dog.

STEP 5: The Tray of bait were put in front of the

two robots. We then observed which tray the dog ate

from first. This was repeated 5 times.

4 RESULTS

4.1 Playmate (Paying Ball)

In Step 1of the playmate experimental procedure, the

care robot threw the ball 10 times. The dogs involved

in the experiment are shown in Table 1, and after the

experiments, the dogs were split into two groups. The

number of the participant dogs that held the ball in

their mouths is nine [C, E, F, G, H, J, L, N, and P],

and this occurred an average of 9.88 times.

The number of the participant dogs that held the

ball in their mouths is eight [A, B, D, I, K, M, O, and

Q], and this occurred an average of 0.375 times.

In the following figures, we refer to the dogs that

held the ball in their mouths as “dogs that play ball

with robots,” and we refer to the dogs group less than

five time that held the ball in their mouths as “dogs

that do not play ball with robots.”

The experiment disclosed that care robot can play

the ball with dogs. And, it is equal human fed.

The bait trays were placed in front of the two

robots in experimental procedure Steps 3 and 5. We

observed which tray the dog ate from first 10 times

each for the two groups previously mentioned. The

results are shown in Figures 8 and 9. The group,

“dogs that play ball with robots,” (nine dogs) chose to

eat the food an average of 6.22 times when it was

placed in front of care robot and 3.66 times when

placed before non-care robot (Figure 8). Dog P did

not eat the food of either robot one time out of the ten;

however, she ate the bait the other nine times. The

group, “dogs that do not play ball with robots,” (eight

dogs) chose to eat the food an average of 4.125 times

when it was placed before care robot and 5.75 times

when the tray was put in front of non-care robot

(Figure 9). Dog D refused to eat the food in front of

both robots once out of ten times; however, she, too,

ate the bait the other nine times.

We calculated the number of selected times “dogs

that play ball with robots” first ate from each robot.

The result is P = 0.000194. The calculation of the t-

test showed that the behavioral changes of the dogs

are statistically significant. The study notes that “dogs

that play ball with robots” chose care robot more

times than non-care robot.

We calculated the number of selected times “dogs

that do not play ball with robots” first ate from each

robot. The result is P = 0.0164. The calculation of the

t-test showed that the behavioral changes of the dogs

are statistically significant. The study suggests that

“dogs that do not play ball with robots” chose non-

care robot more times than care robot.

The video of the experiment is in

https://www.youtube.com/channel/UC8Zri5sVRCU

w1kpesuNByVQ.

Figure 8: The number of times “dogs that play ball with

robots” first ate the bait in front of each robot after playing

ball.

An Observation of Behavioral Changes of Indoor Dogs in Response to Caring Behavior by Humanoid Robots - Can Dogs and Robots Be

Companions?

485

Figure 9: The number of selected times “dogs that do not

play ball with robots” first ate the bait in front of each robot

after playing ball.

4.2 Feeding

In experimental procedure Step 1, the care robot fed

each dog 10 times. The fifteen dogs ate the bait an

average of 7.33 times. The trays were placed in front

of the two robots in experimental procedure Steps 3

and 5. The average results are different from the

frequency of times the human fed the dogs. We

summarize that robots can feed dogs, but the

frequency of times the dogs ate are not equal.

We observed which tray of food the dog ate first

10 times. The tray the dogs chose averaged 6 times

for the care robot and 4 times for the non-care robot

(Figure 10).

We calculated the number of selected times a

robot was chosen by the dogs. The result is P =

0.002224. The calculation of the t-test showed that

these behavioral changes of the dogs are statistically

significant. The results show that dogs chose care

robot more often than non-care robot.

Figure 10: The number of times a dog first selected each

robot after the first feeding.

5 DISCUSSION

5.1 Feeding

“Dogs that play ball with robots” [C, E, F, G, H, J, L,

N, and P] chose the care robot an average of 5.875

times and non-care robot an average of 4.125 times

(Figure 11). Note that “dogs that do not play ball with

robots” [A, B, D, I, K, M, O, and Q] chose the care

robot an average 6.125 times and chose non-care

robot an average of 3.857 times (Figure 12).

In both groups, the dogs chose care robot more

often than non-care robot. In particular, there is a

significant difference between the choices of care

robot and non-core robot in “dogs that do not play ball

with robots.” The result is P = 0.002224. Results

show that “dogs that play ball with robots” chose

eating from care robot more often.

Figure 11: The number of times “dogs that play ball with

robots” first ate the bait in front of each robot after feeding.

Figure 12: The number of selected times “dogs that do not

play ball with robots” first ate the bait in front of each robot

after feeding.

ICAART 2017 - 9th International Conference on Agents and Artiﬁcial Intelligence

486

5.2 Promotion of Positive Interaction

Behavioral guidelines of robots in order to promote

the positive interaction of dogs and robots are:

1) The robot can play ball with “dogs that play ball

with robots.”

2) The robot can feed “dogs that do not play ball

with robots.”

5.3 Prevention of Negative Interaction

In the playing-ball experiment, “dogs that do not play

ball with robots” chose the non-care robot in some

cases. It is considered that “dogs that do not play ball

with robots” have a negative impression of the care

robot. Some dogs interacted negatively with the

robot, but other dogs interacted positively with the

same robot.

We observed a negative interaction in feeding

experiments. A behavior of the robot surprised the

dog, and the dog ran away from robot because the

robot moved suddenly (Figure 13). Behavioral

guidelines of robots hope to prevent future negative

interaction such as this. The most noted guideline is:

1) The robot needs to move slowly when the dog is

near.

Figure 13: The dog was surprised and ran away from robot.

6 RELATED RESEARCHES

The human-robot interaction field has conducted

various studies related to the spread of the robots in

homes: medication management for the elderly at

home (Prakash, 2013), dog-inspired hearing robots

lead participants to sound sources (Koay, 2013),

social robots elicit increased learning in children

(Kennedy, 2015). In relationships between humans

and dogs, dogs avoid people who behave negatively

toward their owner (Chijiiwa, 2015), dogs like

humans who give food to other humans (Kundey,

2011), and there is the oxytocin-gaze positive loop

between a human and dog when they look each other

(Nagasawa, 2015).

In animal-computer interaction, Zeagler et al.

observed a dog’s touchscreen interactions (Zeagler,

2014). They used yellow and blue circles screen, and

dogs touched the circles with their nose. This has

been used to note that dogs can differentiate between

yellow and blue (Neitz, 1989). Mancine et al.

proposed a canine cancer detection system (Mancini,

2015). Baskin et al. observed that dogs can play on

tablets (Baskin, 2015). Robots affected the dogs’

performance by Lakatos et al. (Lakatos, 2014) Many

of these studies observe animal behavior in a system.

In this study, we aimed to build good relationships

between pets and robots at home. We observed dog

behavioral reactions to two types of robots (care robot

and non-care robot). For the dissemination of robots

to the everyday home, we need examine not only

human-robot interaction, but also pet-robot

interaction. Positive pet-robot interaction must also

consider the characteristics of each pet and the

relationship between pet and owner.

7 CONCLUSIONS

This study was conducted as an experiment for

possible the coexistence of pets and robots in the

home. Both a human and a robot took care of the

dogs. The behavior of the dog toward the human, care

robot and, non-care robot was closely monitored. This

study taught us that robots can take care of a dog

instead of a human. We also discovered that care

behavior of a robot influences the behavior of the dog.

In the experiment, we used two robots (care robot and

non-care robot) and placed bait in front of them. We

observed which bait the dog ate first. In the

experiment of playing ball, there were two groups:

“dogs that play ball with robots,” and “dogs that do

not play ball with robots.” Different behavior of the

dogs toward care robot and non-care robot showed it

is better for a robot to throw a ball to promote positive

interaction. In the feeding experiment, we noted

different behaviors of dogs between care robot and

non-care robot. It is better for a robot to feed the dogs

to prevent negative interactions. From each dog’s

reaction to sudden movements during the experiment,

it is best if the robot moves slowly when the dog is

near.

For future experiments, increasing the

participating number of dogs would be useful. We

will conduct various versions of this experiment to

explore other questions. For example, can a dog learn

through playing with the robot? Is the behavior of a

dog different depending on the relationship between

An Observation of Behavioral Changes of Indoor Dogs in Response to Caring Behavior by Humanoid Robots - Can Dogs and Robots Be

Companions?

487

the owner and the robot? Using the results of the this

study, adapting algorithm of the robot’s behavior to

dog, researches on home robot will advance.

ACKNOWLEDGEMENTS

This work was supported by JSPS KAKENHI, Grant

Numbers 26330081, 2687020, and 16K124111.

The authors would like to thank Professor

Isshiki, Associate professor Sugimura, Associate

professor Yamazaki, and the Kanagawa Institute of

Technology for their help in this experiment.

These experiments were conducted with Voice

Research Inc., including Dr. Nishimura and Mr.

Yazawa.

REFERENCES

Akanksha Prakash, Jenay M. Beer, Travis Deyle, Cory-Ann

Smarr, Tiffany L. Chen, Tracy L. Mitzner, Charles C.

Kemp, Wendy A. Rogers, 2013. Older adults'

medication management in the home: How can robots

help?, Human-Robot Interaction (HRI), ACM/IEEE

International Conference, pp.283-290.

Aldenaran, 2015. http://www.softbank.jp/en/robot/

Jay Neitz, Timothy Geist and Gerald H. Jacobs, 1989. Color

vision in the dog, in Visual Neuroscience, 3, pp.119-

125.

Clara Mancini, Rob Harris, Brendan Aengenheister, Claire

Guest, 2015. Re-Centering Multispecies Practices: A

Canine Interface for Cancer Detection Dogs, CHI,

pp.2673-2682.

Clint Zeagler, Scott Gilliland, Larry Freil, Thad Starner and

Melody Moore Jackson, 2014. Going to the

Dogs:Towards an Interactive Touchscreen Interface for

Working Dogs, 27th ACM User Interface Software and

Technology Symposium (UIST), pp.497-507.

Gabriella Lakatos, Mariusz Janiak, Lukasz Malek, Robert

Muszynski, Veronika Konok, Krzysztof Tchon, A.

Miklosi, 2014. Sensing sociality in dogs: what may

make an interactive robot social?, Animal Cognition,

volume 17, Issue 2, pp.387-397.

Hitomi Chijiiwa, Hika Kuroshima, Yusuke Hori, James R.

Anderson and Kazuo Fujita, 2015. Dogs avoid people

who behave negatively to their owner: third-party

affective evaluation, Animal Behaviour, volume 106,

pp.123-127.

James Kennedy, Paul Baxter and Tony Belpaeme, 2015.

The Robot Who Tried Too Hard: Social Behaviour of a

Robot Tutor Can Negatively Affect Child Learning,

ACM/IEEE International Conference on Human-Robot

Interaction, pp.67-74.

K. L. Koay, G. Lakatos, D.S. Syrdal, M. Gácsi, B.

Bereczky, K. Dautenhahn, A. Miklósi and M. L.

Walters, 2013. Hey! There is someone at your door. A

Hearing Robot using Visual Communication Signals of

Hearing Dogs to Communicate Intent, Artificial Life

(ALIFE), IEEE Symposium, pp.90-97.

Kundey SM, De Los Reyes A, Royer E, Molina S, Monnier

B, German R and Coshun A, 2011. Reputation like

inference in domestic dogs (Canis familiaris). Animal

Cognition, pp.291–302.

Miho Nagasawa, Shouhei Mitsui, Shiori En, Nobuyo

Ohtani, Mitsuaki Ohta, Yasuo Sakuma, Tatsushi

Onaka, Kazutaka Mogi and Takefumi Kikusui, 2015.

Oxytocin-gaze positive loop and the coevolution of

human-dog bonds, Science, volume 17, pp.333-336.

Sofya Baskin, Anna Zamansky, 2015. The Player is

Chewing the Tablet!: Towards a Systematic Analysis of

User Behavior in Animal-Computer Interaction, CHI

PLAY, pp.463-468.

ICAART 2017 - 9th International Conference on Agents and Artiﬁcial Intelligence

488