Talk presented at the 2017 Context-oriented programming workshop colocated with ECOOP in barcelona.
This work presents a reinforcement learning based solution to resolve conflicts between dynamic adaptations.
Seal of Good Local Governance (SGLG) 2024Final.pptx
Peace COrP: Learning to solve conflicts between contexts
1. @ncardoz
TICSw
Peace COrP
Nicolás Cardozo Ivana Dusparic Jorge H. Castro
Learning to solve conflicts between
contexts
9th International Workshop on Context-Oriented Programming (COP’17) - Barcelona, Spain
n.cardozo@uniandes.edu.co
16. Learning the best action to a context
16
Action
State
Q-value
1 1
0
-
0
Speed40.activate();
Action
State
Q-value
1 1
0
-
0
Action
State
Q-value
1 1
0
-
0
learnerSpeed
learnerCollision
learnerLane
W-value
(learner with the
largest w-value)
17. Validation
17
Static Resolution Dynamic Resolution Learned Resolution
% of crashes Steps at speed limit Steps on the correct lane1 2 3
h"ps://github.com/jcastrov/Peace-COrP
23. Wrap-up
23
@ncardoz
✓ Adapting to the most appropriate behavior …
….. with respect to the system’s objective
✓ Resolution is dynamic, but converges over time
✓ Reduced development complexity
➡ Composition of adaptations
➡ Use fine-grained contexts to learn behavior
Questions ?