plays “sueca” emys: a social robot that · 2020. 2. 28. · plays “sueca” 1. motivation 2....
TRANSCRIPT
![Page 1: PLAYS “SUECA” EMYS: A SOCIAL ROBOT THAT · 2020. 2. 28. · PLAYS “SUECA” 1. Motivation 2. Goals 3. Background 4. Related Work a. AI in Games b. HRI 5. Proposed Architecture](https://reader035.vdocuments.us/reader035/viewer/2022071015/5fce69d058486b4e7e7bf8ad/html5/thumbnails/1.jpg)
EMYS: A SOCIAL ROBOT THAT PLAYS “SUECA”
![Page 2: PLAYS “SUECA” EMYS: A SOCIAL ROBOT THAT · 2020. 2. 28. · PLAYS “SUECA” 1. Motivation 2. Goals 3. Background 4. Related Work a. AI in Games b. HRI 5. Proposed Architecture](https://reader035.vdocuments.us/reader035/viewer/2022071015/5fce69d058486b4e7e7bf8ad/html5/thumbnails/2.jpg)
1. Motivation2. Goals3. Background4. Related Work
a. AI in Gamesb. HRI
5. Proposed Architecture6. Evaluation7. Conclusion
OUTLINE
![Page 3: PLAYS “SUECA” EMYS: A SOCIAL ROBOT THAT · 2020. 2. 28. · PLAYS “SUECA” 1. Motivation 2. Goals 3. Background 4. Related Work a. AI in Games b. HRI 5. Proposed Architecture](https://reader035.vdocuments.us/reader035/viewer/2022071015/5fce69d058486b4e7e7bf8ad/html5/thumbnails/3.jpg)
1. MOTIVATION
![Page 4: PLAYS “SUECA” EMYS: A SOCIAL ROBOT THAT · 2020. 2. 28. · PLAYS “SUECA” 1. Motivation 2. Goals 3. Background 4. Related Work a. AI in Games b. HRI 5. Proposed Architecture](https://reader035.vdocuments.us/reader035/viewer/2022071015/5fce69d058486b4e7e7bf8ad/html5/thumbnails/4.jpg)
MOTIVATION
![Page 5: PLAYS “SUECA” EMYS: A SOCIAL ROBOT THAT · 2020. 2. 28. · PLAYS “SUECA” 1. Motivation 2. Goals 3. Background 4. Related Work a. AI in Games b. HRI 5. Proposed Architecture](https://reader035.vdocuments.us/reader035/viewer/2022071015/5fce69d058486b4e7e7bf8ad/html5/thumbnails/5.jpg)
MOTIVATION
![Page 6: PLAYS “SUECA” EMYS: A SOCIAL ROBOT THAT · 2020. 2. 28. · PLAYS “SUECA” 1. Motivation 2. Goals 3. Background 4. Related Work a. AI in Games b. HRI 5. Proposed Architecture](https://reader035.vdocuments.us/reader035/viewer/2022071015/5fce69d058486b4e7e7bf8ad/html5/thumbnails/6.jpg)
MOTIVATION
![Page 7: PLAYS “SUECA” EMYS: A SOCIAL ROBOT THAT · 2020. 2. 28. · PLAYS “SUECA” 1. Motivation 2. Goals 3. Background 4. Related Work a. AI in Games b. HRI 5. Proposed Architecture](https://reader035.vdocuments.us/reader035/viewer/2022071015/5fce69d058486b4e7e7bf8ad/html5/thumbnails/7.jpg)
MOTIVATION
![Page 8: PLAYS “SUECA” EMYS: A SOCIAL ROBOT THAT · 2020. 2. 28. · PLAYS “SUECA” 1. Motivation 2. Goals 3. Background 4. Related Work a. AI in Games b. HRI 5. Proposed Architecture](https://reader035.vdocuments.us/reader035/viewer/2022071015/5fce69d058486b4e7e7bf8ad/html5/thumbnails/8.jpg)
2. GOALS
![Page 9: PLAYS “SUECA” EMYS: A SOCIAL ROBOT THAT · 2020. 2. 28. · PLAYS “SUECA” 1. Motivation 2. Goals 3. Background 4. Related Work a. AI in Games b. HRI 5. Proposed Architecture](https://reader035.vdocuments.us/reader035/viewer/2022071015/5fce69d058486b4e7e7bf8ad/html5/thumbnails/9.jpg)
GOALS
Integrate a social robot with aged humans in a card game scenario
● Develop an agent that plays competently Sueca● Develop a socially present embodied agent● Evaluate the correctness of the system
![Page 10: PLAYS “SUECA” EMYS: A SOCIAL ROBOT THAT · 2020. 2. 28. · PLAYS “SUECA” 1. Motivation 2. Goals 3. Background 4. Related Work a. AI in Games b. HRI 5. Proposed Architecture](https://reader035.vdocuments.us/reader035/viewer/2022071015/5fce69d058486b4e7e7bf8ad/html5/thumbnails/10.jpg)
3. BACKGROUND
![Page 11: PLAYS “SUECA” EMYS: A SOCIAL ROBOT THAT · 2020. 2. 28. · PLAYS “SUECA” 1. Motivation 2. Goals 3. Background 4. Related Work a. AI in Games b. HRI 5. Proposed Architecture](https://reader035.vdocuments.us/reader035/viewer/2022071015/5fce69d058486b4e7e7bf8ad/html5/thumbnails/11.jpg)
BACKGROUND
![Page 12: PLAYS “SUECA” EMYS: A SOCIAL ROBOT THAT · 2020. 2. 28. · PLAYS “SUECA” 1. Motivation 2. Goals 3. Background 4. Related Work a. AI in Games b. HRI 5. Proposed Architecture](https://reader035.vdocuments.us/reader035/viewer/2022071015/5fce69d058486b4e7e7bf8ad/html5/thumbnails/12.jpg)
BACKGROUND
Sueca
![Page 13: PLAYS “SUECA” EMYS: A SOCIAL ROBOT THAT · 2020. 2. 28. · PLAYS “SUECA” 1. Motivation 2. Goals 3. Background 4. Related Work a. AI in Games b. HRI 5. Proposed Architecture](https://reader035.vdocuments.us/reader035/viewer/2022071015/5fce69d058486b4e7e7bf8ad/html5/thumbnails/13.jpg)
BACKGROUND
Hidden information?
Information Set!
![Page 14: PLAYS “SUECA” EMYS: A SOCIAL ROBOT THAT · 2020. 2. 28. · PLAYS “SUECA” 1. Motivation 2. Goals 3. Background 4. Related Work a. AI in Games b. HRI 5. Proposed Architecture](https://reader035.vdocuments.us/reader035/viewer/2022071015/5fce69d058486b4e7e7bf8ad/html5/thumbnails/14.jpg)
BACKGROUND
Monte-Carlo Tree Search
1. Selection2. Expansion3. Simulation4. Backpropagation
![Page 15: PLAYS “SUECA” EMYS: A SOCIAL ROBOT THAT · 2020. 2. 28. · PLAYS “SUECA” 1. Motivation 2. Goals 3. Background 4. Related Work a. AI in Games b. HRI 5. Proposed Architecture](https://reader035.vdocuments.us/reader035/viewer/2022071015/5fce69d058486b4e7e7bf8ad/html5/thumbnails/15.jpg)
4. RELATED WORK
![Page 16: PLAYS “SUECA” EMYS: A SOCIAL ROBOT THAT · 2020. 2. 28. · PLAYS “SUECA” 1. Motivation 2. Goals 3. Background 4. Related Work a. AI in Games b. HRI 5. Proposed Architecture](https://reader035.vdocuments.us/reader035/viewer/2022071015/5fce69d058486b4e7e7bf8ad/html5/thumbnails/16.jpg)
4.1. AI IN GAMES
![Page 17: PLAYS “SUECA” EMYS: A SOCIAL ROBOT THAT · 2020. 2. 28. · PLAYS “SUECA” 1. Motivation 2. Goals 3. Background 4. Related Work a. AI in Games b. HRI 5. Proposed Architecture](https://reader035.vdocuments.us/reader035/viewer/2022071015/5fce69d058486b4e7e7bf8ad/html5/thumbnails/17.jpg)
AI IN GAMES
Solving hidden information games...
● Monte-Carlo Methods● Nash-Equilibrium Strategy● Belief distributions
![Page 18: PLAYS “SUECA” EMYS: A SOCIAL ROBOT THAT · 2020. 2. 28. · PLAYS “SUECA” 1. Motivation 2. Goals 3. Background 4. Related Work a. AI in Games b. HRI 5. Proposed Architecture](https://reader035.vdocuments.us/reader035/viewer/2022071015/5fce69d058486b4e7e7bf8ad/html5/thumbnails/18.jpg)
AI IN GAMES
Solving hidden information games...
● Monte-Carlo Methods● Nash-Equilibrium Strategy● Belief distributions
![Page 19: PLAYS “SUECA” EMYS: A SOCIAL ROBOT THAT · 2020. 2. 28. · PLAYS “SUECA” 1. Motivation 2. Goals 3. Background 4. Related Work a. AI in Games b. HRI 5. Proposed Architecture](https://reader035.vdocuments.us/reader035/viewer/2022071015/5fce69d058486b4e7e7bf8ad/html5/thumbnails/19.jpg)
AI IN GAMES
How to deal with hidden information?
PIMC
MONTE-CARLO METHODS
IIMCISMCTS
![Page 20: PLAYS “SUECA” EMYS: A SOCIAL ROBOT THAT · 2020. 2. 28. · PLAYS “SUECA” 1. Motivation 2. Goals 3. Background 4. Related Work a. AI in Games b. HRI 5. Proposed Architecture](https://reader035.vdocuments.us/reader035/viewer/2022071015/5fce69d058486b4e7e7bf8ad/html5/thumbnails/20.jpg)
AI IN GAMES
PIMC
MONTE-CARLO METHODS
Domains Pros / Cons Hidden Information
Bridge
Skat
● Simpler to implement● Strategy fusion● Non-locality
Determinization
![Page 21: PLAYS “SUECA” EMYS: A SOCIAL ROBOT THAT · 2020. 2. 28. · PLAYS “SUECA” 1. Motivation 2. Goals 3. Background 4. Related Work a. AI in Games b. HRI 5. Proposed Architecture](https://reader035.vdocuments.us/reader035/viewer/2022071015/5fce69d058486b4e7e7bf8ad/html5/thumbnails/21.jpg)
AI IN GAMES
ISMCTS
MONTE-CARLO METHODS
Domains Pros / Cons Hidden Information
Dou Dizhu
● Computational Budget● Strategy fusion (less)● Non-locality● Harder to implement
Information Set
![Page 22: PLAYS “SUECA” EMYS: A SOCIAL ROBOT THAT · 2020. 2. 28. · PLAYS “SUECA” 1. Motivation 2. Goals 3. Background 4. Related Work a. AI in Games b. HRI 5. Proposed Architecture](https://reader035.vdocuments.us/reader035/viewer/2022071015/5fce69d058486b4e7e7bf8ad/html5/thumbnails/22.jpg)
AI IN GAMES
IIMC
MONTE-CARLO METHODS
Domains Pros / Cons Hidden Information
Skat
● Player Module● Strategy fusion (less)● Non-locality● Harder to implement
Recursive Determ.
![Page 23: PLAYS “SUECA” EMYS: A SOCIAL ROBOT THAT · 2020. 2. 28. · PLAYS “SUECA” 1. Motivation 2. Goals 3. Background 4. Related Work a. AI in Games b. HRI 5. Proposed Architecture](https://reader035.vdocuments.us/reader035/viewer/2022071015/5fce69d058486b4e7e7bf8ad/html5/thumbnails/23.jpg)
AI IN GAMESBELIEF DISTRIBUTIONS
Domain Technique Goal Suitable
Skat Determine the winning probability of a hand Improve the bidding N
Skat Fastest-cut-first heuristic Order moves Y
Skat Considering similar states equally Reduce tree exploration Y
Skat Calculate the mistake rate of each player Improve the bidding ~
Poker Opponent model Improve MCTS policies Y
![Page 24: PLAYS “SUECA” EMYS: A SOCIAL ROBOT THAT · 2020. 2. 28. · PLAYS “SUECA” 1. Motivation 2. Goals 3. Background 4. Related Work a. AI in Games b. HRI 5. Proposed Architecture](https://reader035.vdocuments.us/reader035/viewer/2022071015/5fce69d058486b4e7e7bf8ad/html5/thumbnails/24.jpg)
4.2. HRI
![Page 25: PLAYS “SUECA” EMYS: A SOCIAL ROBOT THAT · 2020. 2. 28. · PLAYS “SUECA” 1. Motivation 2. Goals 3. Background 4. Related Work a. AI in Games b. HRI 5. Proposed Architecture](https://reader035.vdocuments.us/reader035/viewer/2022071015/5fce69d058486b4e7e7bf8ad/html5/thumbnails/25.jpg)
HRI
Integrate a social robot with aged humans in a card game scenario
Robots in elderly care Social robots in games
![Page 26: PLAYS “SUECA” EMYS: A SOCIAL ROBOT THAT · 2020. 2. 28. · PLAYS “SUECA” 1. Motivation 2. Goals 3. Background 4. Related Work a. AI in Games b. HRI 5. Proposed Architecture](https://reader035.vdocuments.us/reader035/viewer/2022071015/5fce69d058486b4e7e7bf8ad/html5/thumbnails/26.jpg)
HRIROBOTS IN ELDERLY CARE
![Page 27: PLAYS “SUECA” EMYS: A SOCIAL ROBOT THAT · 2020. 2. 28. · PLAYS “SUECA” 1. Motivation 2. Goals 3. Background 4. Related Work a. AI in Games b. HRI 5. Proposed Architecture](https://reader035.vdocuments.us/reader035/viewer/2022071015/5fce69d058486b4e7e7bf8ad/html5/thumbnails/27.jpg)
HRIROBOTS IN ELDERLY CARE
![Page 28: PLAYS “SUECA” EMYS: A SOCIAL ROBOT THAT · 2020. 2. 28. · PLAYS “SUECA” 1. Motivation 2. Goals 3. Background 4. Related Work a. AI in Games b. HRI 5. Proposed Architecture](https://reader035.vdocuments.us/reader035/viewer/2022071015/5fce69d058486b4e7e7bf8ad/html5/thumbnails/28.jpg)
HRIROBOTS IN ELDERLY CARE
![Page 29: PLAYS “SUECA” EMYS: A SOCIAL ROBOT THAT · 2020. 2. 28. · PLAYS “SUECA” 1. Motivation 2. Goals 3. Background 4. Related Work a. AI in Games b. HRI 5. Proposed Architecture](https://reader035.vdocuments.us/reader035/viewer/2022071015/5fce69d058486b4e7e7bf8ad/html5/thumbnails/29.jpg)
HRISOCIAL ROBOTS IN GAMES
● Children tutor● Careful advices● Long-term interactions
![Page 30: PLAYS “SUECA” EMYS: A SOCIAL ROBOT THAT · 2020. 2. 28. · PLAYS “SUECA” 1. Motivation 2. Goals 3. Background 4. Related Work a. AI in Games b. HRI 5. Proposed Architecture](https://reader035.vdocuments.us/reader035/viewer/2022071015/5fce69d058486b4e7e7bf8ad/html5/thumbnails/30.jpg)
HRISOCIAL ROBOTS IN GAMES
Improved social presence:
● Topology of speeches● Relevance value of a move● Power of a player● Simulation of roles● Luck perception
![Page 31: PLAYS “SUECA” EMYS: A SOCIAL ROBOT THAT · 2020. 2. 28. · PLAYS “SUECA” 1. Motivation 2. Goals 3. Background 4. Related Work a. AI in Games b. HRI 5. Proposed Architecture](https://reader035.vdocuments.us/reader035/viewer/2022071015/5fce69d058486b4e7e7bf8ad/html5/thumbnails/31.jpg)
5. PROPOSED ARCHITECTURE
![Page 32: PLAYS “SUECA” EMYS: A SOCIAL ROBOT THAT · 2020. 2. 28. · PLAYS “SUECA” 1. Motivation 2. Goals 3. Background 4. Related Work a. AI in Games b. HRI 5. Proposed Architecture](https://reader035.vdocuments.us/reader035/viewer/2022071015/5fce69d058486b4e7e7bf8ad/html5/thumbnails/32.jpg)
PROPOSED ARCHITECTURE
![Page 33: PLAYS “SUECA” EMYS: A SOCIAL ROBOT THAT · 2020. 2. 28. · PLAYS “SUECA” 1. Motivation 2. Goals 3. Background 4. Related Work a. AI in Games b. HRI 5. Proposed Architecture](https://reader035.vdocuments.us/reader035/viewer/2022071015/5fce69d058486b4e7e7bf8ad/html5/thumbnails/33.jpg)
PROPOSED ARCHITECTUREAI MODULE
● PIMC● Opponent model
■ Cards' predictions■ Actions' predictions
How to collect data?
● Ask for it● Collect it! (it requires
a platform)
![Page 34: PLAYS “SUECA” EMYS: A SOCIAL ROBOT THAT · 2020. 2. 28. · PLAYS “SUECA” 1. Motivation 2. Goals 3. Background 4. Related Work a. AI in Games b. HRI 5. Proposed Architecture](https://reader035.vdocuments.us/reader035/viewer/2022071015/5fce69d058486b4e7e7bf8ad/html5/thumbnails/34.jpg)
6. EVALUATION
![Page 35: PLAYS “SUECA” EMYS: A SOCIAL ROBOT THAT · 2020. 2. 28. · PLAYS “SUECA” 1. Motivation 2. Goals 3. Background 4. Related Work a. AI in Games b. HRI 5. Proposed Architecture](https://reader035.vdocuments.us/reader035/viewer/2022071015/5fce69d058486b4e7e7bf8ad/html5/thumbnails/35.jpg)
EVALUATION
Develop an agent that plays competently Sueca
Performance measures● Game points● Offline pre-computation time
● These measures will be compared to different parametrizations and a naive approach
● University community will test it
![Page 36: PLAYS “SUECA” EMYS: A SOCIAL ROBOT THAT · 2020. 2. 28. · PLAYS “SUECA” 1. Motivation 2. Goals 3. Background 4. Related Work a. AI in Games b. HRI 5. Proposed Architecture](https://reader035.vdocuments.us/reader035/viewer/2022071015/5fce69d058486b4e7e7bf8ad/html5/thumbnails/36.jpg)
EVALUATION
Develop a socially present embodied agent
Two conditions● Few or nonexisting social behaviours● Several behaviours from the game state
● The elderly will test it● Godspeed - participants’ perception of the robot● Networked Minds - presence perception
![Page 37: PLAYS “SUECA” EMYS: A SOCIAL ROBOT THAT · 2020. 2. 28. · PLAYS “SUECA” 1. Motivation 2. Goals 3. Background 4. Related Work a. AI in Games b. HRI 5. Proposed Architecture](https://reader035.vdocuments.us/reader035/viewer/2022071015/5fce69d058486b4e7e7bf8ad/html5/thumbnails/37.jpg)
THANK YOU!