I am currently working on implementing a Multi-Agent System for Smart Grids.

There's a lot of literature for that and some things confuse me. I have read that there is FIPA, which aimed to create a unified Agent Communication Language. So multiple Agents are talking to each other and FIPA specifies how the messages should be sent and processed. However, it is pretty old.

In newer papers, where Multi-Agent Reinforcement Learning Algorithms are proposed, FIPA or generally any ACL isn't mentioned. I believe that is because in MARL, communication is done by observing the states of the other agents, rather than communicating explicitly. Also in MARL, the decision making is not based on negotiation like in FIPA but with the learned policy.

I am now super confused if I got it right. Is FIPA still a thing I should worry about when I design my Multi-Agent System? Is there any other thing to handle communication in MARL other than sharing states?

Any help would be really appreciated, thank you very much :)

  • 1
    $\begingroup$ This is a very interesting question. If you found any answer, can you share it with us? $\endgroup$
    – Erik
    Jul 1 at 11:19
  • $\begingroup$ So I don't know how correct my thoughts are, if somebody knows better I am open to suggestions. However my understanding has changed that there are several aspects to Multi-Agent Systems. In most AI related fields, communication is not the research focus and most things get done in simulations only, whereas in software engineering, communication protocols like FIPA or ACL are the main interest. However I didn't find lots of new literature using FIPA, my guess is that it either is used in already existing frameworks (like JADE) or other protocols are used (there is new tech since 2006 :D). $\endgroup$ Jul 5 at 7:55

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.