1. Designed by Gusz Eiben & Mark Hoogendoorn
On-line adaptation, learning,
evolution
2. Designed by Gusz Eiben & Mark Hoogendoorn
Outline
• Population-based Adaptive Systems
• Types of adaptation: evolution, individual
(lifetime) learning, social learning
• Machine learning
• Reinforcement learning
• Off-line vs. on-line adaptation
3. Designed by Gusz Eiben & Mark Hoogendoorn
Population-based Adaptive Systems
PAS have two essential features
•They consist of a group of basic units that can
perform actions, e.g., computation,
communication, interaction, etc.
•The ability to adapt at
– individual level (modify agent ) and/or
– group level (add/remove agent).
4. Designed by Gusz Eiben & Mark Hoogendoorn
Types of adaptation
• Evolutionary learning (EL): Changes at population
level (assumed non-Lamarckian)
• Lifetime learning (LL): Changes at agent level
– Individual learning (IL): adaptation autonomously
through a purely internal procedure
– Social learning (SL): adaptation through interaction
/communication
5. Designed by Gusz Eiben & Mark Hoogendoorn
Taxonomy of adaptation
Adaptation
Evolutionary
Learning
Lifetime
Learning
Individual
Learning
Social
Learning
6. Designed by Gusz Eiben & Mark Hoogendoorn
Taxonomy of adaptation 2
Adaptation
Evolutionary
Learning
Lifetime
Learning
Individual
Learning
Social
Learning
Learning
Evolution
7. Designed by Gusz Eiben & Mark Hoogendoorn
Adaptation ≠ operation
• Operation: controller is being used
– Sensory inputs outputs (motor, comm. device)
– Robot behavior changes, not the controller
• Adaptation: controller is being changed
– Present controller new controller
– Uses utility/reward/fitness info
– It may require
• One single robot – learning
• More robots – evolution, social learning
• Adaptation + operation = generate + test
• Off-line (initial controller design, before start) vs. on-line (after
start)
8. Designed by Gusz Eiben & Mark Hoogendoorn
Genotype
Developmental
Engine(decoder)
Genetic operators:
mutation & xover
Learning
operators
Robot
behavior
State of the
environment
Phenotype =
controller
Reward
Fitness
Selection
operators
9. Designed by Gusz Eiben & Mark Hoogendoorn
Genotype
Developmental
Engine(decoder)
Genetic operators:
mutation & xover
Learning
operators
Robot
behavior
State of the
environment
Phenotype =
controller
Reward
Fitness
Selection
operators
10. Designed by Gusz Eiben & Mark Hoogendoorn
Genotype
Developmental
Engine(decoder)
Genetic operators:
mutation & xover
Learning
operators
Robot
behavior
State of the
environment
Reward
Fitness
Selection
operators
Phenotype
controllershape
11. Designed by Gusz Eiben & Mark Hoogendoorn
Phenotype
Genotype
Developmental
Engine(decoder)
Genetic operators:
mutation & xover
Learning
operators
Robot
behavior
State of the
environment
Reward
Fitness
Selection
operators
controllershape
12. Designed by Gusz Eiben & Mark Hoogendoorn
Evolutionary loop
Genotype
DevelopmentalEngine
Genetic operators:
mutation & xover
Learning operator(s)
Robot
behavior
Changes in
environment
Controller =
phenotype
Reward
Fitness
Selection
operator(s)
13. Designed by Gusz Eiben & Mark Hoogendoorn
Learning loop
Genotype
DevelopmentalEngine
Genetic operators:
mutation & xover
Learning operator(s)
Robot
behavior
Changes in
environment
Controller =
phenotype
Reward
Fitness
Selection
operator(s)
14. Designed by Gusz Eiben & Mark Hoogendoorn
ENVIRONMENTAGENT
Reward r(t)
State s(t)
Action a(t)
15. Designed by Gusz Eiben & Mark Hoogendoorn
Reinforcement learning
Agent in situation/state st chooses action at
World changes to situation/state st+1
Agent perceives situation st+1 and gets reward rt+1
Telling the agent what to do is its
POLICY πt(s, a) = P r{at = a|st = s}
Given the situation at time t is s, the policy gives the probability the agent’s
action will be a.
For example: πt(s, goforward) = 0.5, πt(s, gobackward) = 0.5.
Reinforcement learning ⇒ Get/find/learn the policy
16. Designed by Gusz Eiben & Mark Hoogendoorn
Further reading
• Evert Haasdijk and A.E. Eiben and Alan F.T.
Winfield, Individual Social and Evolutionary
Adaptation in Collective Systems , Serge
Kernbach (eds.) , Handbook of Collective
Robotics , Pan Stanford , 2011