SlideShare a Scribd company logo
1 of 20
Download to read offline
Adaptation
                          in embodied & situated agents
                                      Author: Claudio Martella
                          Collaborators: Dott. Stefano Nolfi (ISTC - CNR)
                               Prof. N.A. Borghese (AIS Lab - UniMi)

                                          October, 2011


                                                1
Tuesday, October 11, 11
The problem
                It is difficult to build autonomous systems
                through a top-down approach:


                     • the behavior might be too complex for the
                          designer to control
                     • the environment is noisy and not perfect
                     • the world is unpredictable
                                                2
Tuesday, October 11, 11
Evolutionary robotics is a branch of robotics
                       that uses evolutionary methodologies
                  to develop controllers for autonomous robots.

                                        Nolfi, Floreano [2004] - MIT Press




                                        3
Tuesday, October 11, 11
The objective

                            We wanted to analyze the possibility
                                of applying adaptive processes
                               to embodied & situated agents
                                          considering
                          evolutionary, individual and social learning.



                                                4
Tuesday, October 11, 11
E&S agents

                     • Embodied: the agent can exploit the
                          characteristics of the robot (shape,
                          sensors, actuators etc.).
                     • Situated: the solution can exploit the
                          possible interactions that the environments
                          offers.


                                              5
Tuesday, October 11, 11
The methodology
                          E-puck Robot       Simulation




            Problem: categorize 10 objects (Good, Poisonous)
                                         6
Tuesday, October 11, 11
The evolutionary process




                                     7
Tuesday, October 11, 11
1st goal
                   Implement an algorithm for individual learning.

                                  The algorithm should start
                            with one set of candidate parameters
                          and it would modify them by trial & error.

                          Decision: start from Simulated Annealing *


          * "Optimization by Simulated Annealing", Kirkpatrick, S.; Gelatt, C. D.; Vecchi, M. P. (1983) - Science
                                                             8
Tuesday, October 11, 11
Simulated Annealing
                                       Temperature:

                                       It probabilistically accepts
                                        mutations that decrease
                                               the fitness.

                                       The probability decreases
                                              with time.

                                       It allows the algorithm to
                                       jump out of local minima.
                                   9
Tuesday, October 11, 11
Stochasticity in E&S
                                Evaluation depends on
                            the (random) initial conditions:




                                           10
Tuesday, October 11, 11
The intuition
                            Temperature                                  Stochasticity

             0.9                                         0.9

         0.675                                         0.675

           0.45                                         0.45

         0.225                                         0.225

                0                                         0
                    100   200   300   400   500                10   20       30      40   50

         Probability of accepting negative             Probability of accepting negative
          mutations decreases with the                  mutations decreases with the
                 increase of time                         increase of #evaluations
                                                  11
Tuesday, October 11, 11
Contributions
             Substitute external stochasticity with internal:

             •       Remove Temperature

             •       Start with few evaluations and increase with time


                                       Results
             •       Simplifies the algorithm

             •       Better performance (~10% improvement)

             •       Lighter algorithm (~50% less evaluations for us)

                                               12
Tuesday, October 11, 11
2nd goal
                          Implement an algorithm for social learning.

                            The algorithm should take advantage
                           of the interaction with an expert agent
                               to acquire an adaptive solution
                            that is improved and/or in less time.

                   Decision: apply individual learning to imitation.

                                              13
Tuesday, October 11, 11
Why?

                 Social learning should avoid reinventing the wheel.

                 In principle, when guided, learning is faster & safer.

                          It should be the basis for cultural evolution.




                                                14
Tuesday, October 11, 11
How?
               There are simpler forms of social learning:


                • social facilitation
                • contagious behavior
                • stimulus enhancement

                                       15
Tuesday, October 11, 11
How (technically)?




           Fitness function: student should learn to give
           outputs similar to the agent’s, given the same input.
                                     16
Tuesday, October 11, 11
How (technically)?
               Pure imitation brings to under-fitting individuals.
                      We introduced a hybrid approach.


    f it = f itsoc · (1      ↵) + f itind · ↵

                 c
    ↵=           N




                                          17
Tuesday, October 11, 11
Contributions
             •       Modeled social learning with simple form of imitation

             •       Modeled hybrid social-individual learning approach


                                       Results
             •       Performance on the problem is not improved

             •       Adaptive behavior is acquired faster

             •       More agents acquire an adaptive behavior

                                               18
Tuesday, October 11, 11
Intuitive interpretation
          parameters space                         solutions space




                                  Social learning as a method
                          for promising initial parameters selection.
                                  Social learning as a method
                               for jumping out of local maxima.
                                              19
Tuesday, October 11, 11
Questions?



                              20
Tuesday, October 11, 11

More Related Content

Viewers also liked

Representantes comenius1
Representantes comenius1Representantes comenius1
Representantes comenius1Rosafcid
 
Hadoop: A Hands-on Introduction
Hadoop: A Hands-on IntroductionHadoop: A Hands-on Introduction
Hadoop: A Hands-on IntroductionClaudio Martella
 
Greek trivia2
Greek trivia2Greek trivia2
Greek trivia2sbillia
 
Relay for life
Relay for lifeRelay for life
Relay for lifesbillia
 
Disney trivia2
Disney trivia2Disney trivia2
Disney trivia2sbillia
 
Greek trivia
Greek triviaGreek trivia
Greek triviasbillia
 
Pro sports trivia
Pro sports triviaPro sports trivia
Pro sports triviasbillia
 
Giraph at Hadoop Summit 2014
Giraph at Hadoop Summit 2014Giraph at Hadoop Summit 2014
Giraph at Hadoop Summit 2014Claudio Martella
 

Viewers also liked (10)

Laurel's slide show
Laurel's slide showLaurel's slide show
Laurel's slide show
 
Haikus 1
Haikus 1Haikus 1
Haikus 1
 
Representantes comenius1
Representantes comenius1Representantes comenius1
Representantes comenius1
 
Hadoop: A Hands-on Introduction
Hadoop: A Hands-on IntroductionHadoop: A Hands-on Introduction
Hadoop: A Hands-on Introduction
 
Greek trivia2
Greek trivia2Greek trivia2
Greek trivia2
 
Relay for life
Relay for lifeRelay for life
Relay for life
 
Disney trivia2
Disney trivia2Disney trivia2
Disney trivia2
 
Greek trivia
Greek triviaGreek trivia
Greek trivia
 
Pro sports trivia
Pro sports triviaPro sports trivia
Pro sports trivia
 
Giraph at Hadoop Summit 2014
Giraph at Hadoop Summit 2014Giraph at Hadoop Summit 2014
Giraph at Hadoop Summit 2014
 

Similar to Adaptation in Embodied & Situated Agents

Similar to Adaptation in Embodied & Situated Agents (6)

VBPR 1st seminar
VBPR 1st seminarVBPR 1st seminar
VBPR 1st seminar
 
Why do a designed experiment
Why do a designed experimentWhy do a designed experiment
Why do a designed experiment
 
Adapt or Go extinct
Adapt or Go extinctAdapt or Go extinct
Adapt or Go extinct
 
When Tdd Goes Awry
When Tdd Goes AwryWhen Tdd Goes Awry
When Tdd Goes Awry
 
A Cloud Outage Under the Lens of “Profound Knowledge”
A Cloud Outage Under the Lens of  “Profound Knowledge” A Cloud Outage Under the Lens of  “Profound Knowledge”
A Cloud Outage Under the Lens of “Profound Knowledge”
 
Exploratory Testing in an Agile Context
Exploratory Testing in an Agile ContextExploratory Testing in an Agile Context
Exploratory Testing in an Agile Context
 

Recently uploaded

Tata AIG General Insurance Company - Insurer Innovation Award 2024
Tata AIG General Insurance Company - Insurer Innovation Award 2024Tata AIG General Insurance Company - Insurer Innovation Award 2024
Tata AIG General Insurance Company - Insurer Innovation Award 2024The Digital Insurer
 
🐬 The future of MySQL is Postgres 🐘
🐬  The future of MySQL is Postgres   🐘🐬  The future of MySQL is Postgres   🐘
🐬 The future of MySQL is Postgres 🐘RTylerCroy
 
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024The Digital Insurer
 
Boost Fertility New Invention Ups Success Rates.pdf
Boost Fertility New Invention Ups Success Rates.pdfBoost Fertility New Invention Ups Success Rates.pdf
Boost Fertility New Invention Ups Success Rates.pdfsudhanshuwaghmare1
 
Manulife - Insurer Innovation Award 2024
Manulife - Insurer Innovation Award 2024Manulife - Insurer Innovation Award 2024
Manulife - Insurer Innovation Award 2024The Digital Insurer
 
Understanding Discord NSFW Servers A Guide for Responsible Users.pdf
Understanding Discord NSFW Servers A Guide for Responsible Users.pdfUnderstanding Discord NSFW Servers A Guide for Responsible Users.pdf
Understanding Discord NSFW Servers A Guide for Responsible Users.pdfUK Journal
 
Top 5 Benefits OF Using Muvi Live Paywall For Live Streams
Top 5 Benefits OF Using Muvi Live Paywall For Live StreamsTop 5 Benefits OF Using Muvi Live Paywall For Live Streams
Top 5 Benefits OF Using Muvi Live Paywall For Live StreamsRoshan Dwivedi
 
AWS Community Day CPH - Three problems of Terraform
AWS Community Day CPH - Three problems of TerraformAWS Community Day CPH - Three problems of Terraform
AWS Community Day CPH - Three problems of TerraformAndrey Devyatkin
 
Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers:  A Deep Dive into Serverless Spatial Data and FMECloud Frontiers:  A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FMESafe Software
 
Strategies for Landing an Oracle DBA Job as a Fresher
Strategies for Landing an Oracle DBA Job as a FresherStrategies for Landing an Oracle DBA Job as a Fresher
Strategies for Landing an Oracle DBA Job as a FresherRemote DBA Services
 
Repurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost Saving
Repurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost SavingRepurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost Saving
Repurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost SavingEdi Saputra
 
Apidays New York 2024 - The Good, the Bad and the Governed by David O'Neill, ...
Apidays New York 2024 - The Good, the Bad and the Governed by David O'Neill, ...Apidays New York 2024 - The Good, the Bad and the Governed by David O'Neill, ...
Apidays New York 2024 - The Good, the Bad and the Governed by David O'Neill, ...apidays
 
Strategize a Smooth Tenant-to-tenant Migration and Copilot Takeoff
Strategize a Smooth Tenant-to-tenant Migration and Copilot TakeoffStrategize a Smooth Tenant-to-tenant Migration and Copilot Takeoff
Strategize a Smooth Tenant-to-tenant Migration and Copilot Takeoffsammart93
 
Connector Corner: Accelerate revenue generation using UiPath API-centric busi...
Connector Corner: Accelerate revenue generation using UiPath API-centric busi...Connector Corner: Accelerate revenue generation using UiPath API-centric busi...
Connector Corner: Accelerate revenue generation using UiPath API-centric busi...DianaGray10
 
Artificial Intelligence Chap.5 : Uncertainty
Artificial Intelligence Chap.5 : UncertaintyArtificial Intelligence Chap.5 : Uncertainty
Artificial Intelligence Chap.5 : UncertaintyKhushali Kathiriya
 
Why Teams call analytics are critical to your entire business
Why Teams call analytics are critical to your entire businessWhy Teams call analytics are critical to your entire business
Why Teams call analytics are critical to your entire businesspanagenda
 
Polkadot JAM Slides - Token2049 - By Dr. Gavin Wood
Polkadot JAM Slides - Token2049 - By Dr. Gavin WoodPolkadot JAM Slides - Token2049 - By Dr. Gavin Wood
Polkadot JAM Slides - Token2049 - By Dr. Gavin WoodJuan lago vázquez
 
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...apidays
 
A Year of the Servo Reboot: Where Are We Now?
A Year of the Servo Reboot: Where Are We Now?A Year of the Servo Reboot: Where Are We Now?
A Year of the Servo Reboot: Where Are We Now?Igalia
 
The 7 Things I Know About Cyber Security After 25 Years | April 2024
The 7 Things I Know About Cyber Security After 25 Years | April 2024The 7 Things I Know About Cyber Security After 25 Years | April 2024
The 7 Things I Know About Cyber Security After 25 Years | April 2024Rafal Los
 

Recently uploaded (20)

Tata AIG General Insurance Company - Insurer Innovation Award 2024
Tata AIG General Insurance Company - Insurer Innovation Award 2024Tata AIG General Insurance Company - Insurer Innovation Award 2024
Tata AIG General Insurance Company - Insurer Innovation Award 2024
 
🐬 The future of MySQL is Postgres 🐘
🐬  The future of MySQL is Postgres   🐘🐬  The future of MySQL is Postgres   🐘
🐬 The future of MySQL is Postgres 🐘
 
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
 
Boost Fertility New Invention Ups Success Rates.pdf
Boost Fertility New Invention Ups Success Rates.pdfBoost Fertility New Invention Ups Success Rates.pdf
Boost Fertility New Invention Ups Success Rates.pdf
 
Manulife - Insurer Innovation Award 2024
Manulife - Insurer Innovation Award 2024Manulife - Insurer Innovation Award 2024
Manulife - Insurer Innovation Award 2024
 
Understanding Discord NSFW Servers A Guide for Responsible Users.pdf
Understanding Discord NSFW Servers A Guide for Responsible Users.pdfUnderstanding Discord NSFW Servers A Guide for Responsible Users.pdf
Understanding Discord NSFW Servers A Guide for Responsible Users.pdf
 
Top 5 Benefits OF Using Muvi Live Paywall For Live Streams
Top 5 Benefits OF Using Muvi Live Paywall For Live StreamsTop 5 Benefits OF Using Muvi Live Paywall For Live Streams
Top 5 Benefits OF Using Muvi Live Paywall For Live Streams
 
AWS Community Day CPH - Three problems of Terraform
AWS Community Day CPH - Three problems of TerraformAWS Community Day CPH - Three problems of Terraform
AWS Community Day CPH - Three problems of Terraform
 
Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers:  A Deep Dive into Serverless Spatial Data and FMECloud Frontiers:  A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FME
 
Strategies for Landing an Oracle DBA Job as a Fresher
Strategies for Landing an Oracle DBA Job as a FresherStrategies for Landing an Oracle DBA Job as a Fresher
Strategies for Landing an Oracle DBA Job as a Fresher
 
Repurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost Saving
Repurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost SavingRepurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost Saving
Repurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost Saving
 
Apidays New York 2024 - The Good, the Bad and the Governed by David O'Neill, ...
Apidays New York 2024 - The Good, the Bad and the Governed by David O'Neill, ...Apidays New York 2024 - The Good, the Bad and the Governed by David O'Neill, ...
Apidays New York 2024 - The Good, the Bad and the Governed by David O'Neill, ...
 
Strategize a Smooth Tenant-to-tenant Migration and Copilot Takeoff
Strategize a Smooth Tenant-to-tenant Migration and Copilot TakeoffStrategize a Smooth Tenant-to-tenant Migration and Copilot Takeoff
Strategize a Smooth Tenant-to-tenant Migration and Copilot Takeoff
 
Connector Corner: Accelerate revenue generation using UiPath API-centric busi...
Connector Corner: Accelerate revenue generation using UiPath API-centric busi...Connector Corner: Accelerate revenue generation using UiPath API-centric busi...
Connector Corner: Accelerate revenue generation using UiPath API-centric busi...
 
Artificial Intelligence Chap.5 : Uncertainty
Artificial Intelligence Chap.5 : UncertaintyArtificial Intelligence Chap.5 : Uncertainty
Artificial Intelligence Chap.5 : Uncertainty
 
Why Teams call analytics are critical to your entire business
Why Teams call analytics are critical to your entire businessWhy Teams call analytics are critical to your entire business
Why Teams call analytics are critical to your entire business
 
Polkadot JAM Slides - Token2049 - By Dr. Gavin Wood
Polkadot JAM Slides - Token2049 - By Dr. Gavin WoodPolkadot JAM Slides - Token2049 - By Dr. Gavin Wood
Polkadot JAM Slides - Token2049 - By Dr. Gavin Wood
 
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
 
A Year of the Servo Reboot: Where Are We Now?
A Year of the Servo Reboot: Where Are We Now?A Year of the Servo Reboot: Where Are We Now?
A Year of the Servo Reboot: Where Are We Now?
 
The 7 Things I Know About Cyber Security After 25 Years | April 2024
The 7 Things I Know About Cyber Security After 25 Years | April 2024The 7 Things I Know About Cyber Security After 25 Years | April 2024
The 7 Things I Know About Cyber Security After 25 Years | April 2024
 

Adaptation in Embodied & Situated Agents

  • 1. Adaptation in embodied & situated agents Author: Claudio Martella Collaborators: Dott. Stefano Nolfi (ISTC - CNR) Prof. N.A. Borghese (AIS Lab - UniMi) October, 2011 1 Tuesday, October 11, 11
  • 2. The problem It is difficult to build autonomous systems through a top-down approach: • the behavior might be too complex for the designer to control • the environment is noisy and not perfect • the world is unpredictable 2 Tuesday, October 11, 11
  • 3. Evolutionary robotics is a branch of robotics that uses evolutionary methodologies to develop controllers for autonomous robots. Nolfi, Floreano [2004] - MIT Press 3 Tuesday, October 11, 11
  • 4. The objective We wanted to analyze the possibility of applying adaptive processes to embodied & situated agents considering evolutionary, individual and social learning. 4 Tuesday, October 11, 11
  • 5. E&S agents • Embodied: the agent can exploit the characteristics of the robot (shape, sensors, actuators etc.). • Situated: the solution can exploit the possible interactions that the environments offers. 5 Tuesday, October 11, 11
  • 6. The methodology E-puck Robot Simulation Problem: categorize 10 objects (Good, Poisonous) 6 Tuesday, October 11, 11
  • 7. The evolutionary process 7 Tuesday, October 11, 11
  • 8. 1st goal Implement an algorithm for individual learning. The algorithm should start with one set of candidate parameters and it would modify them by trial & error. Decision: start from Simulated Annealing * * "Optimization by Simulated Annealing", Kirkpatrick, S.; Gelatt, C. D.; Vecchi, M. P. (1983) - Science 8 Tuesday, October 11, 11
  • 9. Simulated Annealing Temperature: It probabilistically accepts mutations that decrease the fitness. The probability decreases with time. It allows the algorithm to jump out of local minima. 9 Tuesday, October 11, 11
  • 10. Stochasticity in E&S Evaluation depends on the (random) initial conditions: 10 Tuesday, October 11, 11
  • 11. The intuition Temperature Stochasticity 0.9 0.9 0.675 0.675 0.45 0.45 0.225 0.225 0 0 100 200 300 400 500 10 20 30 40 50 Probability of accepting negative Probability of accepting negative mutations decreases with the mutations decreases with the increase of time increase of #evaluations 11 Tuesday, October 11, 11
  • 12. Contributions Substitute external stochasticity with internal: • Remove Temperature • Start with few evaluations and increase with time Results • Simplifies the algorithm • Better performance (~10% improvement) • Lighter algorithm (~50% less evaluations for us) 12 Tuesday, October 11, 11
  • 13. 2nd goal Implement an algorithm for social learning. The algorithm should take advantage of the interaction with an expert agent to acquire an adaptive solution that is improved and/or in less time. Decision: apply individual learning to imitation. 13 Tuesday, October 11, 11
  • 14. Why? Social learning should avoid reinventing the wheel. In principle, when guided, learning is faster & safer. It should be the basis for cultural evolution. 14 Tuesday, October 11, 11
  • 15. How? There are simpler forms of social learning: • social facilitation • contagious behavior • stimulus enhancement 15 Tuesday, October 11, 11
  • 16. How (technically)? Fitness function: student should learn to give outputs similar to the agent’s, given the same input. 16 Tuesday, October 11, 11
  • 17. How (technically)? Pure imitation brings to under-fitting individuals. We introduced a hybrid approach. f it = f itsoc · (1 ↵) + f itind · ↵ c ↵= N 17 Tuesday, October 11, 11
  • 18. Contributions • Modeled social learning with simple form of imitation • Modeled hybrid social-individual learning approach Results • Performance on the problem is not improved • Adaptive behavior is acquired faster • More agents acquire an adaptive behavior 18 Tuesday, October 11, 11
  • 19. Intuitive interpretation parameters space solutions space Social learning as a method for promising initial parameters selection. Social learning as a method for jumping out of local maxima. 19 Tuesday, October 11, 11
  • 20. Questions? 20 Tuesday, October 11, 11