hpi-sam / Robust-Multi-Agent-Reinforcement-Learning-for-SAS

Research project on robust multi-agent reinforcement learning (marl) for self-adaptive systems (sas)
MIT License
0 stars 0 forks source link

Issue are remaining in the model #32

Open christianadriano opened 2 years ago

christianadriano commented 2 years ago

Concerning the message "Issue are remaining in the model!". Sona, just answered me. Follows. [Chris] We noticed that the errors are happening when we apply ReplaceComponent rule. Is there something particular about it? [Sona] Yes, that rule is only applicable for components for which an alternative component type is available for replacement, which is only the authentication service component. Should the replacement rule be applied to any other component type, the rule is ineffective as it cannot replace the faulty component, thus, error remains in the system. [Chris] So, it would make sense to filter out this rule for any component except the authentication service. Currently, this rule is being available to the other components.

So, my suggestion is that filter-out this rule for any component != "AuthenticationService". Ideally, this filtering should happen before mRubis sends the list of available actions. Could you please give a look how to do that?