[robocup-rescue-s] Centralized Decisions

Hi folks!

Thanks Cameron Skinner for your detailed answer!
I have another question:
Starting from my old question:
> 4) Can I center all strategy in a external class ? For example: All communication from all platoon comes to
> this peace of code and they decide what to do ?
> The only restriction on code is that agents are not allowed to share
>their world model. For example, in Java you can't have something like:
> public class MyAgentClass {
> private static WorldModel modelForAllAgents;
>...
>}
> because all instances of MyAgentClass will use the same WorldModel
>object. The idea is that we're trying to model the real world and having
>shared state like this is not realistic.
>There is nothing stopping you having external classes for performing
>computation, communication or anything else. Just make sure you don't
>share state between agents.

Can I center all status information of buildings, road, platoon in a single class?
All I want is the humanoids such as fire brigade and police will sense and send a message using tell command to this centralized peace of code and then it will decide based on priorities, location and last actions.
Can I do that ? Is it a good idea or I should let the agents autonomous?

Thanks!

Adolfo Ferreira

----------------------------------------
> Date: Sun, 17 Jun 2007 11:55:02 +1200
> From: cam_at_cs.auckland.ac.nz
> To: adolfo.ferreira_at_hotmail.com
> CC: robocup-rescue-s_at_cc.gatech.edu
> Subject: Re: [robocup-rescue-s] Starter Questions
>
> Hi.
>
> Adolfo Bravo Ferreira wrote:
> >
> > Hi folks,
> >
> > Im starting with YAB API using NetBeans and I will ask you a *few* questions:
> >
> > 1) Does a Fire Station / Ambulance center / Police Office / Refuge gets burned ? If it does will I loose the same score as a normal building?
>
> This is an option in the fire simulator. I think by default they do not
> burn, but you probably shouldn't rely on this.
>
> > 2) Whats the purpose of a Fire Station / Ambulance center / Police Office / River ? Only to inform ?
>
> The three centres are basically for communication. In versions before
> 0.50 (and 0.49.9+) communication could only occur between a platoon
> agent (e.g. fire brigade) and other platoon agents of the same type and
> with the corresponding centre. Centres could communicate with each
> other, too. So, for a fire brigade to send a message to an ambulance
> team the message had to go:
>
> Fire brigade -> Fire station -> Ambulance centre -> Ambulance team
>
> With the new communication architecture this restriction is lifted, but
> in the future we will probably make the communication more realistic.
> For example, platoon agents might only be allowed to listen to 1 channel
> but centres could listen to several channels.
>
> Rivers are not used at the moment.
>
> > 3) Does AmbulanceTeam / FireBrigade / PoliceForce / car gets burned too ? Or only Civilians? If my rescue team gets burned I need to bring them to a Refuge to restore their hp ?
>
> Platoon agents can get burned and trapped, just like civilians.
> Returning to a refuge will remove damage, but not restore lost HP.
> Damage is the number of HP lost each turn due to injury.
>
> For example, my fire brigade has 1000HP and a damage of 50. Next
> timestep it will have 950HP, then 900 and so on. If it returns to a
> refuge then it will have 900HP and 0 damage so the HP will stop
> decreasing. If it gets damaged again in the future then the HP value
> will start decreasing again. When HP=0 the agent is dead.
>
> Cars are not used at the moment.
>
> > 4) Can I center all strategy in a external class ? For example: All communication from all platoon comes to this peace of code and they decide what to do ?
>
> The only restriction on code is that agents are not allowed to share
> their world model. For example, in Java you can't have something like:
>
> public class MyAgentClass {
> private static WorldModel modelForAllAgents;
> ...
> }
>
> because all instances of MyAgentClass will use the same WorldModel
> object. The idea is that we're trying to model the real world and having
> shared state like this is not realistic.
>
> There is nothing stopping you having external classes for performing
> computation, communication or anything else. Just make sure you don't
> share state between agents.
>
> > 6) Have any range for communication ? Agents can sense only 30m from their location. Does communication has any range ? An agent to communicate to another agent has any range limit ?
>
> Voice messages have a range of 30m, agents can see only 10m. Radio
> messages have infinite range, but only agents that are listening to the
> right channel can hear them. Voice messages are assigned channel number 0.
>
> > 7) Do I need to work with communication ? Can I compress this communication to store more data in the limit size?
>
> Yes, and you probably should. Generally an agent gets more information
> in a SENSE message than it can transmit via TELL messages, so you'll
> need something to ensure that all your agents share relevant information
> and coordinate effectively.
>
> > 8) All the strategy does not need to run in the same time as the cycles right ?
>
> No, you can implement it in any way you wish. Remember that the kernel
> will only accept commands from agents for a short time after each sense
> message, so make sure that you send your commands quickly after
> receiving the sense message.
>
> I hope that helps.
>
> Regards,
> Cameron Skinner.

_________________________________________________________________
Express yourself instantly with MSN Messenger! Download today it's FREE!
http://messenger.msn.click-url.com/go/onm00200471ave/direct/01/
_______________________________________________
robocup-rescue-s mailing list
robocup-rescue-s_at_cc.gatech.edu
https://lists.cc.gatech.edu/mailman/listinfo/robocup-rescue-s
Received on Sun 17 Jun 2007 - 14:47:06 GMT