Modeling Human Imitation as Probabilistic Inference and an Application in Robotics

A part of the invited talk, "Infant imitation, brain, and social cognition" by Andrew N. Meltzoff, at HRI 2015 "Cognition: A Bridge between Robotics and Interaction" Workshop (http://www.macs.hw.ac.uk/~kl360/HRI2015W/index.html)

Transcript

1. The “Like-Me” Hypothesis Self-experience allows infants to interpret the act of other Self-experience plays an important role in goal inference and imitation Meltzoff, Dev Sci, 2007
 Meltzoff, Acta Psychologica, 2007 Computational Model: Probabilistic instantiation of “Like-me” hypothesis
2. Computational Model : goal : action : start state : final state Friesen & Rao, Cog Sci, 2011
3. Computational Model: Gaze Following : desired fixation location : head motor commands : start head pose : final head pose Friesen & Rao, Cog Sci, 2011
4. Computational Model: Gaze Following : desired fixation location : head motor commands : start head pose : final head pose Random Variables are Continuous Relations are Probabilistic Functions
5. Probabilistic Functions: Gaussian Processes (GP) ‣ nonparametric ‣ probabilistic distribution over functions Effective with small numbers of training examples
6. Modeling Relations Transition model: Learned from “Exploration” or “Body Babbling”
7. Modeling Relations Policy model: Learned from “Rejection sampling” “Planning via Probabilistic Inference” (Verma & Rao, NIPS, 2005)
8. Goal-directed Planning G A B Xi Xf Given goal and start state, infer action. Closed form solutions.
9. Goal Inference G A B Xi Xf Given start state and final state, infer goal. Approximate via Bayesian Monte Carlo.
10. Gaze Following Gm Am Bm Xm i Xm f A G B Xi X f change of coords. Given: mentor start state, mentor final state, start state
 Infer: mentor goal, action OTHER SELF Infer goal
11. Model Simulation 0 200 400 600 −600 −400 −200 0 200 400 x position (cm) y position (cm) Agent Mentor true fixation points inferred fixation points 500 -500 500 100
12. Model Performance 0 200 400 600 −600 −400 −200 0 200 400 x position (cm) y position (cm) Agent Mentor true fixation points inferred fixation points
13. Blindfold Self-Experience Task Gm Am Bm Xm i Xm f A G B Xi X f change of coords.
14. Blindfold Self-Experience Task Baseline Window Opaque 0 1 2 3 4 Looking Score (a) Computational Model (b) Infant Data AGENT (simulation results) INFANT (actual behavior data) Friesen & Rao, Cog Sci, 2011
15. Computational Model: Robotic Tabletop Manipulation : desired tabletop position : control commands : start object location : final object location Random Variables are Discrete Relations are Probability Tables
16. Robotic Tabletop Manipulation Task Chung et al, Tech report, 2014
17. Learning: Transition Model Transition model:
18. Learning: Policy Model Policy model:
19. Goal-Based Imitation and Action Selection Step1: Goal-Inference Step II: Action-Inference Step III:
 Final State Prediction Gm Am Bm Xm i Xm f A G B Xi X f change of coords.
20. Learned Transition Model Place Right Place Off-table Push Right Push Off-table Place Right Place Off-table Push Right Push Off-table
21. Learned Transition Model Place Right Place Off-table Push Right Push Off-table Place Right Place Off-table Push Right Push Off-table
22. Learned Transition Model Place Right Place Off-table Push Right Push Off-table Place Right Place Off-table Push Right Push Off-table
23. Goal Inference Goal ✴MAP Prediction
24. Action Inference Place Left Place Right Place Off Push Left Push Right Push Off Place Left Place Right Place Off Push Left Push Right Push Off Goal ✴MAP Prediction Demonstration: Xi = LEFT, Xf = RIGHT
25. Final State Prediction
26. Imitation Results
27. A Bayesian Developmental Approach to Goal-Based Imitation Learning Michael Jae-Yoon Chung, Abram L. Friesen,
 Dieter Fox, Andrew N. Meltzoff, and Rajesh P.N. Rao University of Washington HRI2015 Workshop Cognition: A Bridge between robotics and interaction
28. Imitation Learning in Humans Not mere trajectory following. Imitation based on goal inference. Demonstration Goal-based imitation Infants above 1.5 years of age can imitate action even from an unsuccessful demonstration (Meltzoff & Brook 1998)
29. Another Example: Gaze Following and Blindfolds 12 month olds Meltzoff & Brooks, Dev Psych Blind fold experience no training after training
30. The “Like-Me” Hypothesis Self-experience allows infants to interpret the act of others. Self-experience plays an important role in goal inference and imitation. Computational Model: Probabilistic instantiation of “Like-me” hypothesis. Meltzoff, Dev Sci, 2007 Metlzoff, Acta Psychologica, 2007
31. Model Performance 0 10 20 30 40 0 0.05 0.1 0.15 0.2 gaze error (degrees) p(error) −40 −20 0 0 0.05 0.1 0.15 0.2 gaze error (degrees) p(error) −40 −20 0 0 0.05 0.1 0.15 0.2 gaze error (degrees) p(error) (a) Forward Inference (b) Reverse Inference (c) Gaze Following