私は、交通信号制御機を (マスター) エージェントとして表し、相互に通信するプロジェクトに取り組んでいます。これにはJADEを使おうと思っていたのですが、AKKAに出会いました。AKKA について読んだところ、マスターが 1 つあり、マスターが最初にワーカーを作成することがわかりました。その場合、(すべてのマスター) エージェント表現を実装するにはどうすればよいですか?
1206 次
1 に答える
2
So where'd you end up on this? What I've determined is that the actor model is not the same thing as the agent model, and that, in order to make agents make sense in AKKA, you need to do quite a bit yourself.
because AKKA is going to send all requests through a router, you need to store state in the message itself.
in order to emulate the masterless behavior of agent peers, you will have to build a routing master that will know which agents to send what to. Also, if you have a "repeated averaging" problem, like group opinions, that master will need to watch over the state for the problem / send the results of other agents to each agent on the next iteration.
I think you may need to ask what you're looking for in your framework. If you're looking for ruggedness, AKKA is probably an OK bet (it's inspired by ErLang), if you're looking for an agent platform that can run simulations, JADE or RePast is probably a much better bet.
At least for agent systems, AKKA is about as much of a match as ActiveMQ or ZeroMQ is (i.e. it handles some communication problems for distributed systems, but it's not an agent based framework).
于 2015-04-13T02:25:50.773 に答える