Passive Demonstrations of Light-Based Robot Signals for Improved - - PowerPoint PPT Presentation

passive demonstrations of light based robot signals for
SMART_READER_LITE
LIVE PREVIEW

Passive Demonstrations of Light-Based Robot Signals for Improved - - PowerPoint PPT Presentation

Justin W. Hart Passive Demonstrations of Light-Based Robot Signals for Improved Human Interpretability Rolando Fernandez, Nathan John, Sean Kirmani, Justin Hart, Jivko Sinapov, and Peter Stone Building-Wide Intelligence Project Learning


slide-1
SLIDE 1

Passive Demonstrations of Light-Based Robot Signals for Improved Human Interpretability

Rolando Fernandez, Nathan John, Sean Kirmani, Justin Hart, Jivko Sinapov, and Peter Stone

Building-Wide Intelligence Project Learning Agents Research Group Department of Computer Science The University of Texas at Austin

Justin W. Hart

slide-2
SLIDE 2
  • Real-world deployment
  • Fleet of autonomous

service robots

– 5 (soon to be 7!)

  • Respond to verbal

commands

  • Intended to provide

services to the building’s

  • ccupants

Building-Wide Intelligence

Justin W. Hart

slide-3
SLIDE 3
  • The robots navigate crowded hallways
  • This can lead to navigational conflicts

where the robot and person want to walk into the same place.

Navigating Crowded Spaces

Justin W. Hart

slide-4
SLIDE 4

The Hallway Problem

Justin W. Hart

slide-5
SLIDE 5

Left Turn Right Turn

Solution: LED Turn Signals

Justin W. Hart

slide-6
SLIDE 6

Hallway Test Environment

Justin W. Hart

slide-7
SLIDE 7

Modeling the Hallway – 3 Traffic Lanes

Justin W. Hart

slide-8
SLIDE 8
  • Hallway – 17x1.85 meters
  • Signal – 7 meters from the person
  • Execute – 2.75 meters from the person
  • Conflict – 1 meter from the person

Robot Behavior

Justin W. Hart

slide-9
SLIDE 9
  • 13 participants (9 male, 4 female)
  • Recruited primarily from the UT Austin Engineering and

Computer Science.

  • Procedure
  • Informed consent & media release
  • Participants walked once down the test hallway with the robot

going the opposite direction

  • Survey
  • Results
  • Half interpreted the signal the direction they should take.
  • Half interpreted the signal like a turn signal.

Pilot Study

Justin W. Hart

slide-10
SLIDE 10
  • Want to demonstrate the signal with no explicit training period.
  • Idea – Passive Demonstration
  • Have the robot use the signal in context before it is explicitly

needed in the interaction with the participant.

  • Robot makes a lane change at the start of the hallway.

New Approach – Passive Demonstration

Justin W. Hart

slide-11
SLIDE 11

Study Conditjons Passive Demonstratjon No Demonstratjon Demonstratjon LED Signal LED No LED No Demonstratjon, No LED Demonstratjon, No LED LED No Demonstratjon, LED Demonstratjon, LED

Study Design

Justin W. Hart

slide-12
SLIDE 12
  • 47 Participants (39 male, 8 female)
  • Recruited primarily from the UT Austin Engineering and

Computer Science.

  • Procedure
  • Informed consent & media release
  • Participants walked once down the test hallway with the robot

going the opposite direction

  • Survey
  • Additional 11 participants (8 male, 3 female) recruited for a

follow-up condition.

Study

Justin W. Hart

slide-13
SLIDE 13

No Demo, No LED

Justin W. Hart

slide-14
SLIDE 14

Justin W. Hart

slide-15
SLIDE 15

No Demo, LED

Justin W. Hart

slide-16
SLIDE 16

Justin W. Hart

slide-17
SLIDE 17

Demo, No LED

Justin W. Hart

slide-18
SLIDE 18

Justin W. Hart

slide-19
SLIDE 19

Demo, LED

Justin W. Hart

slide-20
SLIDE 20

Justin W. Hart

slide-21
SLIDE 21

Passive Demonstrations Reduce Conflict

Justin W. Hart

slide-22
SLIDE 22
  • We thought the meaning of the LED turn signals would be
  • bvious to users.

– It was not.

  • Introduction of the concept of the passive demonstration of the

signal.

– Effective in demonstrating the meaning of the signal to users.

Conclusion

Justin W. Hart