resilient machines through continuous self-modeling pattern recognition 2010.04.06 seung-hyun lee...
TRANSCRIPT
![Page 1: Resilient Machines Through Continuous Self-Modeling Pattern Recognition 2010.04.06 Seung-Hyun Lee Soft Computing Lab. Josh Bongard,Victor Zykov, and Hod](https://reader036.vdocuments.site/reader036/viewer/2022062800/56649dff5503460f94ae78b3/html5/thumbnails/1.jpg)
Resilient Machines ThroughContinuous Self-Modeling
Pattern Recognition
2010.04.06
Seung-Hyun Lee
Soft Computing Lab.
Josh Bongard,Victor Zykov, and Hod Lipson, Science, Vol.314, pp. 1118-1121, 2006.
![Page 2: Resilient Machines Through Continuous Self-Modeling Pattern Recognition 2010.04.06 Seung-Hyun Lee Soft Computing Lab. Josh Bongard,Victor Zykov, and Hod](https://reader036.vdocuments.site/reader036/viewer/2022062800/56649dff5503460f94ae78b3/html5/thumbnails/2.jpg)
S FT COMPUTING @ YONSEI UNIV . KOREA16
Contents
• Introduction
• Motivation
• Self Modeling
• Experiments
• Conclusion
2 / 15
![Page 3: Resilient Machines Through Continuous Self-Modeling Pattern Recognition 2010.04.06 Seung-Hyun Lee Soft Computing Lab. Josh Bongard,Victor Zykov, and Hod](https://reader036.vdocuments.site/reader036/viewer/2022062800/56649dff5503460f94ae78b3/html5/thumbnails/3.jpg)
S FT COMPUTING @ YONSEI UNIV . KOREA16
Introduction
• Animals– After injured,
create qualitatively different
compensatory behaviors
• Robots– How robots can deal with this sort of unexpected damage?
self modeling
3 / 15
![Page 4: Resilient Machines Through Continuous Self-Modeling Pattern Recognition 2010.04.06 Seung-Hyun Lee Soft Computing Lab. Josh Bongard,Victor Zykov, and Hod](https://reader036.vdocuments.site/reader036/viewer/2022062800/56649dff5503460f94ae78b3/html5/thumbnails/4.jpg)
S FT COMPUTING @ YONSEI UNIV . KOREA16
Motivation
• How can robot learn its own morphology?– Direct observation?– Database of past experience?
• How can robot synthesize complex behaviors or recover from damage?
– Trial and error? slow, costly, risky!
• In this paper,– Inferring morphology: self-directed exploration– Complex behavior or recovering from damage: synthesize new be-
haviors using the resulting self models
4 / 15
![Page 5: Resilient Machines Through Continuous Self-Modeling Pattern Recognition 2010.04.06 Seung-Hyun Lee Soft Computing Lab. Josh Bongard,Victor Zykov, and Hod](https://reader036.vdocuments.site/reader036/viewer/2022062800/56649dff5503460f94ae78b3/html5/thumbnails/5.jpg)
S FT COMPUTING @ YONSEI UNIV . KOREA16
Self Modeling
5 / 15
Overall Process
ModelingPrediction
Testing
![Page 6: Resilient Machines Through Continuous Self-Modeling Pattern Recognition 2010.04.06 Seung-Hyun Lee Soft Computing Lab. Josh Bongard,Victor Zykov, and Hod](https://reader036.vdocuments.site/reader036/viewer/2022062800/56649dff5503460f94ae78b3/html5/thumbnails/6.jpg)
S FT COMPUTING @ YONSEI UNIV . KOREA16
Self Modeling
6 / 15
Testing
• In this process– Performs an arbitrary motor action
– Records the resulting sensory data
![Page 7: Resilient Machines Through Continuous Self-Modeling Pattern Recognition 2010.04.06 Seung-Hyun Lee Soft Computing Lab. Josh Bongard,Victor Zykov, and Hod](https://reader036.vdocuments.site/reader036/viewer/2022062800/56649dff5503460f94ae78b3/html5/thumbnails/7.jpg)
S FT COMPUTING @ YONSEI UNIV . KOREA16
Self Modeling
7 / 15
Modeiling
• Model synthesize component – Synthesizes a set of candidate self-models
• Method– Before damage(topological modeling)
• Greedy random-mutation hill climber algorithm• 16 parameters
Robot initially knows how many body pars it is composed of, the size, weight and mass of each part, and angle-movement relations
• 15 random models• 200 iterations• Evaluation:
Euclidean distance between the centroid and where the centroid should be
– After damage(parametric modeling)• Self-model is frozen• 8 parameters (volumes and masses are scaled by 10%~200%)
![Page 8: Resilient Machines Through Continuous Self-Modeling Pattern Recognition 2010.04.06 Seung-Hyun Lee Soft Computing Lab. Josh Bongard,Victor Zykov, and Hod](https://reader036.vdocuments.site/reader036/viewer/2022062800/56649dff5503460f94ae78b3/html5/thumbnails/8.jpg)
S FT COMPUTING @ YONSEI UNIV . KOREA16
Self Modeling
8 / 15
Prediction
• Action synthesize component– Find a new action most likely to elicit the most information from the
robot based on the current self model inferred
![Page 9: Resilient Machines Through Continuous Self-Modeling Pattern Recognition 2010.04.06 Seung-Hyun Lee Soft Computing Lab. Josh Bongard,Victor Zykov, and Hod](https://reader036.vdocuments.site/reader036/viewer/2022062800/56649dff5503460f94ae78b3/html5/thumbnails/9.jpg)
S FT COMPUTING @ YONSEI UNIV . KOREA16
Self Modeling
9 / 15
• After self modeling procedures(16 times repetition)– Create desired behaviors (D)– Execute by the physical robot
![Page 10: Resilient Machines Through Continuous Self-Modeling Pattern Recognition 2010.04.06 Seung-Hyun Lee Soft Computing Lab. Josh Bongard,Victor Zykov, and Hod](https://reader036.vdocuments.site/reader036/viewer/2022062800/56649dff5503460f94ae78b3/html5/thumbnails/10.jpg)
S FT COMPUTING @ YONSEI UNIV . KOREA16
Self Modeling
10 / 15
![Page 11: Resilient Machines Through Continuous Self-Modeling Pattern Recognition 2010.04.06 Seung-Hyun Lee Soft Computing Lab. Josh Bongard,Victor Zykov, and Hod](https://reader036.vdocuments.site/reader036/viewer/2022062800/56649dff5503460f94ae78b3/html5/thumbnails/11.jpg)
S FT COMPUTING @ YONSEI UNIV . KOREA16
Experiments
• Speculation– 4 upper and lower leg parts and a main body– 8 motorized joints(-90 ~ 90 degree range)
• 0 degree: flat• Positive degree: upwards• Negative degree: downwards
– 2 tilt sensors
• Self model representation– Planar topological arrangement
• Damage– Disabled one leg
11 / 15
Robot
![Page 12: Resilient Machines Through Continuous Self-Modeling Pattern Recognition 2010.04.06 Seung-Hyun Lee Soft Computing Lab. Josh Bongard,Victor Zykov, and Hod](https://reader036.vdocuments.site/reader036/viewer/2022062800/56649dff5503460f94ae78b3/html5/thumbnails/12.jpg)
S FT COMPUTING @ YONSEI UNIV . KOREA16
Experiments
• Control variables– Computational efforts(250,000 internal model simulations)– Physical actions(16)
• Three algorithms– Algorithm 1:
16 random physical actions batch training(modeling)– Algorithm 2:
Physical actions self modeling random action selection– Algorithm 3(proposed):
Physical actions self modeling actions selection
12 / 15
Design
![Page 13: Resilient Machines Through Continuous Self-Modeling Pattern Recognition 2010.04.06 Seung-Hyun Lee Soft Computing Lab. Josh Bongard,Victor Zykov, and Hod](https://reader036.vdocuments.site/reader036/viewer/2022062800/56649dff5503460f94ae78b3/html5/thumbnails/13.jpg)
S FT COMPUTING @ YONSEI UNIV . KOREA16
Experiments
13 / 15
Result
![Page 14: Resilient Machines Through Continuous Self-Modeling Pattern Recognition 2010.04.06 Seung-Hyun Lee Soft Computing Lab. Josh Bongard,Victor Zykov, and Hod](https://reader036.vdocuments.site/reader036/viewer/2022062800/56649dff5503460f94ae78b3/html5/thumbnails/14.jpg)
S FT COMPUTING @ YONSEI UNIV . KOREA16
Experiments
14 / 15
Result
Model-driven algorithm is more accurate than ran-dom baseline algorithms
A robot that actively chooses action on the basis of its current set of hypothesized self-models has a bet-ter chance of successfully inferring its own morphol-ogy
![Page 15: Resilient Machines Through Continuous Self-Modeling Pattern Recognition 2010.04.06 Seung-Hyun Lee Soft Computing Lab. Josh Bongard,Victor Zykov, and Hod](https://reader036.vdocuments.site/reader036/viewer/2022062800/56649dff5503460f94ae78b3/html5/thumbnails/15.jpg)
S FT COMPUTING @ YONSEI UNIV . KOREA16
Experiments
15 / 15
Result
Automatically generated self-model was sufficiently predictive to allow the robot to consistently develop forward motion patterns without further physical tri-als
![Page 16: Resilient Machines Through Continuous Self-Modeling Pattern Recognition 2010.04.06 Seung-Hyun Lee Soft Computing Lab. Josh Bongard,Victor Zykov, and Hod](https://reader036.vdocuments.site/reader036/viewer/2022062800/56649dff5503460f94ae78b3/html5/thumbnails/16.jpg)
S FT COMPUTING @ YONSEI UNIV . KOREA16
Conclusion
• Contribution– First physical system
• Autonomously recover its own morphology with little prior knowledge• Optimize the parameters of its morphology after unexpected change
– Show the possibility of unknown cognitive process• Which organisms actively create and update self models in the brain?• How and which sensor-motor signals are used to do this?• What form these model take?• Does human utilize multiple competing models?
16 / 15
Result
![Page 17: Resilient Machines Through Continuous Self-Modeling Pattern Recognition 2010.04.06 Seung-Hyun Lee Soft Computing Lab. Josh Bongard,Victor Zykov, and Hod](https://reader036.vdocuments.site/reader036/viewer/2022062800/56649dff5503460f94ae78b3/html5/thumbnails/17.jpg)
Thank you