1) AUPAIR is a personal home service robot developed by Seoul National University that won the RoboCup@Home League competition.
2) The RoboCup@Home competition involves various scenarios testing abilities like speech recognition, navigation, and assisting humans.
3) AUPAIR uses perception modules like object detection and activity recognition combined with action modules like following and guiding to solve tasks in the competition scenarios.
AUPAIR: The Personal Home Service Robot that Won the RoboCup@Home League
1. AUPAIR:
The Personal Home Service Robot that
Won the RoboCup@Home League
Beom-Jin Lee, Jinyoung Choi, Chung-Yeon Lee, Kyung-Wha Par
k, Sungjun Choi, Cheolho Han, Dong-Sig Han,
Christina Baek, Patrick Emaase and Byoung-Tak Zhang
Seoul National University, Seoul, Korea
Biointelligence Lab
Team Leader: Beom-Jin Lee
4. Latest RoboCup
• 20th Annual Competition
• Total 5 Leagues
• RoboCupSoccer
• RoboCupIndustrial
• RoboCupRescue
• RoboCup@Home
• RoboCupJunior
• Total 17 Sub-leagues
• More than 3,500 dedicated scientists
• Developers from more than 40 countries
12. RoboCup@Home
• Second largest league in RoboCup
• Initiated the RoboCup@Home competitions in 2005
• Develop service and assistive robot technology for
autonomous service robots
Adaptive
Smart HRI
Wisspeintner, Thomas, et al. "RoboCup@
Home: Scientific competition and
benchmarking for domestic service
robots." Interaction Studies 10.3 (2009):
392-426.
14. RoboCup@Home SSPL Challenges
6 Scenarios + 2 Open Demonstration = 8 Challenges
1. Speech Recognition and Person Recognition
2. Cocktail Party
3. Help Me Carry
4. General Purpose Service Robot (GPSR)
5. Restaurant
6. Tour Guide
7. Enhanced Endurance General Purpose Service Robot (E2GPSR)
8. Open Challenge
9. Final Demonstration
15. RoboCup@Home SSPL Scenarios
Speech Recognition and Person Recognition
1. Play riddle game
2. Play bluff man game
Focus technique
Person, object, speech, activity
recognition
Q&A
Sound localization
12
people
16. RoboCup@Home SSPL Scenarios
Cocktail Party
1. Enter the party room
2. Find 3 standing or sitting guest
3. Go to barman and give order
4. Find missing beverage
5. Correct order
Focus technique
Person, object, speech, activity
recognition
Person description
1
2
3
4
5
17. RoboCup@Home SSPL Scenarios
Help Me Carry
1. Memorize operator and
follow
2. Save car waypoint and find
other person to help him
carry grocery
3. Guide person to car waypoint
Focus technique
Following, online mapping,
guiding
1
2
3
18. RoboCup@Home SSPL Scenarios
General Purpose Service Robot (GPSR)
1. Generate random command
from command generator
2. Solve the task and come back
to starting point
3. Repeat 3 times
Focus technique
Overall skill
19. RoboCup@Home SSPL Scenarios
Restaurant
1. No map
2. Check barman
3. Get order from a person
4. Go to barman and give order
5. Repeat 3-4
Focus technique
Online mapping
20. RoboCup@Home SSPL Scenarios
Tour Guide
1. Go out side
2. Find tourist who needs tour
3. Guide person to bench and
give introduction about
RoboCup
Focus technique
Online mapping
Q&A
1
2
3
21. RoboCup@Home SSPL Scenarios
Enhanced Endurance General Purpose Service Robot (E2GPSR)
Test special skills
1. Three in one (solve three
command within 5 minute)
2. Q&A with GPSR
Focus technique
Navigation
Q&A
Recognition skills
24. RoboCup@Home SSPL
Implemented Modules (Perception)
Human Pose Estimation3Object Detection and Recognition2
[2] Redmon, et al., "YOLO9000: better, faster, stronger." arXiv preprint arXiv:1612.08242 (2016).
[3] Cao, et al. "Realtime multi-person 2d pose estimation using part affinity fields." arXiv preprint arXiv:1611.08050 (2016).
25. RoboCup@Home SSPL
Implemented Modules (Perception)
Human Re-identification4 Scene Description5
[4] Ahmed, et al., "An improved deep learning architecture for person re-identification." Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2015.
[5] Johnson, et al., "Densecap: Fully convolutional localization networks for dense captioning." Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2016.
49. Special Thanks
Kyung-Min Kim Jin-Hwa Kim
Kim, Kyung-Min, et al.
"DeepStory: Video Story
QA by Deep Embedded
Memory Networks." arXiv
preprint arXiv:1707.00836
(2017).)
Kim, Jin-Hwa, et al.
"Multimodal residual
learning for visual
qa." Advances in Neural
Information Processing
Systems. 2016.