1 of 21

Tail Wagging the Dog:

Perceptions of Canine vs Non-Canine Behavior in a Quadruped Robot

Ruchi Bhalani, Alekhya Kuchimanchi, Hanaa Siddiqui, Justin Hart

Learning Agents Research Group (LARG)

2 of 21

Motivation

Quadruped robots have recently become available to researchers and developers ¹
These robots are often referred to as “robot dogs”
Can we behaviorally increase the impression of these robots as being canine?

¹https://www.bostondynamics.com/products/spot

Alekhya

Quadruped robots have recently become available to researchers and are often referred to as “robot dogs”
These robots provide an exciting opportunity to leverage interactions with a dog-like robot rather than a human-like robot
Dogs have a special position in human society, as the first species to be domesticated^[1] and one that has a robust set of social skills for interacting with people
As a starting point and a building block for human-robot interaction in the canine form-factor, this work builds a set of behaviors chosen to encourage people to perceive quadruped robots as though they are dog-like.
In this study, we work with the Boston Dynamics Spot as our quadruped mobile robot.
This research seeks to enhance study participants' perceptions of the robot as being ``dog-like'' by developing a set of robot behaviors reminiscent of canine behavior.

3 of 21

Approach

01

02

03

Determine what constitutes canine-like behavior

Human Survey

Pick out the top responses from the survey to implement on the robot

Analyze

Develop canine body movements on the robot

Body Language

4 of 21

Approach

04

05

Evaluate robot’s new social behaviors

Human Survey

Determine if behavior is more canine.

Evaluate people’s perceptions of the robot.

Analyze

5 of 21

Pre-Survey: What constitutes “canine-like”?

“What behaviors make dogs seem _______?”

Brief video of Boston Dynamics Spot²
What behaviors should be added to the Spot robot in order for it to appear more dog-like?
57 voluntary participants recruited on social media

Friendly Happy Aggressive	Anxious Sad Relaxed

²https://www.youtube.com/watch?v=lwaVAkI5-zk

6 of 21

Number of Participant Mentions Per Behavior

7 of 21

Behaviors Implemented

Tail Wagging

The robot tilts to the right, returns to the center, tilts to the left, returns to the center, and continues in a smoothly-interpolated motion

Play Bow

Like a dog “bowing” during play, robot tilts body forward

8 of 21

Behaviors Implemented

3. Sit

Opposite of play bow, robot tilts body backwards

4. Walk in a Circle

Robot follows linearly interpolated waypoints in the shape of a circle

5. Spin

Mimicking chasing a tail

Like walk in circle but robot turns in a circle in place

9 of 21

Human Survey: “How canine-like does Spot seem?”

Survey Questions adapted from Godspeed Questionnaire³

³C. Bartneck, D. Kulic, E. Croft, and S. Zoghbi, “Measurement instruments for the anthropomorphism, animacy, likeability, perceived intelligence, and perceived safety of robots,” International Journal of Social Robotics, vol. 1, pp. 71–81, November 2009.

Machinelike vs Doglike Fake vs Natural Unconscious vs Conscious Artificial vs Lifelike Remote controlled vs Autonomous Moving Rigidly vs Moving Elegantly	Inert vs Interactive Apathetic vs Responsive Incompetent vs Competent Ignorant vs Knowledgeable Dislike vs Like Unfriendly vs Friendly Unpleasant vs Pleasant Awful vs Nice

Participants rated the robot on a 7-Point Likert scale

10 of 21

Human Survey: “How canine-like does Spot seem?”

Two vignettes:

Canine condition
Non-canine condition

Inter-participant
Each vignette ran on Speedway with a QR code nearby on two separate days
Responses:

112 Canine responses
110 Non-canine responses

11 of 21

https://youtu.be/f2sfL1cJ4PE https://youtu.be/R9vWQJpmh-0

12 of 21

Participants’ Perceptions of the Robot

13 of 21

Conclusion and Applications

The results affirms the effectiveness of the implemented behaviors.

Create enriching interactions
Increase the acceptance of these robots in their deployments

Service dogs
Delivery bots
Exploratory robots

14 of 21

Thank You!

15 of 21

Body Position	Tail Wagging	Play Bow	Sitting
Roll	-π/16 to π/16	0	0
Pitch	0	0 to 3π/14	-π/7 to 0
Yaw	-π/8 to π/8	0	0

Changes in Roll, Pitch, Yaw of Body Positions

16 of 21

Body Movement	Velocityx (m/s)	Velocity y (m/s)	Rotation Angle (rad)	Trajectory Points
Walk in Circle	2	0	-1.5 to 1.5	12
Spin	0	0	-1 to 1	8

Changes in Velocities and Angles of Body Movements

Walk in circle

Spin

17 of 21

18 of 21

Human Survey: “How canine-like does Spot seem?”

When I encountered the robot I felt “___”

Curious
Cautious
Calm, Nervous
Excited
Scared

I would be comfortable seeing this robot “_____”

Walking in the classroom
Walking in the library
Providing delivery services on campus
Providing delivery services to my home, serving food in a restaurant

19 of 21

20 of 21

21 of 21

Conclusion

The addition of canine social body language was significant

Second round of testing with descriptions of behavior

Develop more motions to demonstrate canine social body language