Hi Mahesh, welcome to Akka.
If you have multiple devices, you are right to think about this with multiple actors. Think about those actors as proxies for the internal device state of each device. To whatever degree your system must react to the changes in those devices, your proxies will be the first to detect the changes in each device. And for each of those device types, there is an individual FSM that allows you to track them.
What I think you are also asking is to do something with that information, You don't have enough information here to give strong guidance, but the theory is the same. Once you have proxies for all the devices in the form of actors in a system, it is a far easier modeling job for additional actors to orchestrate a view. If the proxy actors understand the state of individual devices, these "orchestration actors" can show the overall network. Imagining there is a UI, the network graph would talk to the orchestration actor to draw the graph, talking to each proxy actor for information like uptime, other devices they are connected to, etc, but at the level of the proxy, not by talking to the device when that graph needs to be drawn.
Spend time understanding supervision and monitoring. I mentioned this idea a network graph UI. Imagine the whole idea of that graph is live updates and that many people could watch the same graph on their screens with different reasons for links between devices to be highlighted (one user might want a flashing link when the network is down, another might want to see a flashing link when data is moving across it). This is the correct use for another actor, one for each user view, and by using monitoring, the view actor can start with basic information about any changes in existence of a proxy for that view. Monitoring at the Akka level does not send state change events (that's for you to work out), but does deal with the existential problem of whether an actor has suddenly died.
That should get you started...
Brian