Erlang (programming language)/Tutorials/Agents: Difference between revisions
imported>Eric Evers |
imported>Meg Taylor m (spelling: lenght -> length) |
||
(2 intermediate revisions by 2 users not shown) | |||
Line 11: | Line 11: | ||
=====Explanation of the code in jungle.erl===== | =====Explanation of the code in jungle.erl===== | ||
Here we have a simple chat-bot agent called person/4. We create two instances of it called Tarzan and Jane. They talk to each other. Each has a timeout. The timeout is the | Here we have a simple chat-bot agent called person/4. We create two instances of it called Tarzan and Jane. They talk to each other. Each has a timeout. The timeout is the length of time one will wait before they intiate conversation. The initial timeout of Jane is set to 10 seconds. The initial timeout of Tarzan is set to 8 seconds. Because of the initial values, Tarzan will speak first and Jane will respond. Both timeouts start over but keep the same values. Again, Tarzan speaks first and Jane responds. Now things get interesting. The agent can tell if the conversation is repeating. If the conversation repeats then special messages are sent to cause a swap in the relative levels of timeout. Now Tarzan waits longer than Jane and Jane has a chance to speak first. Now, Jane speaks first twice. Then they swap initiative again. Since the processes are autonomous, we need to stop them both with a quit program called jungle:quit(). | ||
Note about timeouts: The agents behave as if in a normal conversation. It is considered rude if one person always starts each conversation because the topic is set at the start. If agent-1 starts the converstion twice in a row then agent-1 should abdicate the starting position by changing the timesouts. An ethernet collision is another situation where timeouts must change to smooth communication. Actually when one agent always speaks first it is the opposite of a collsion, so we might call such a dialog situation an anticollision. If one agent always speaks first then it is more like a traditional client-server situation. | Note about timeouts: The agents behave as if in a normal conversation. It is considered rude if one person always starts each conversation because the topic is set at the start. If agent-1 starts the converstion twice in a row then agent-1 should abdicate the starting position by changing the timesouts. An [[ethernet]] collision is another situation where timeouts must change to smooth communication. Actually when one agent always speaks first it is the opposite of a collsion, so we might call such a dialog situation an anticollision. If one agent always speaks first then it is more like a traditional client-server situation. To cure an anticollision, we must swap the relative magnitude of the timeouts. | ||
=====Example program listing: jungle.erl===== | =====Example program listing: jungle.erl===== |
Latest revision as of 23:55, 8 February 2010
The metadata subpage is missing. You can start it via filling in this form or by following the instructions that come up after clicking on the [show] link to the right. | |||
---|---|---|---|
|
Autonomous Agents in Erlang
Agent Example programs
Two Agent systems
Dynamic timeout based initiative switching
Explanation of the code in jungle.erl
Here we have a simple chat-bot agent called person/4. We create two instances of it called Tarzan and Jane. They talk to each other. Each has a timeout. The timeout is the length of time one will wait before they intiate conversation. The initial timeout of Jane is set to 10 seconds. The initial timeout of Tarzan is set to 8 seconds. Because of the initial values, Tarzan will speak first and Jane will respond. Both timeouts start over but keep the same values. Again, Tarzan speaks first and Jane responds. Now things get interesting. The agent can tell if the conversation is repeating. If the conversation repeats then special messages are sent to cause a swap in the relative levels of timeout. Now Tarzan waits longer than Jane and Jane has a chance to speak first. Now, Jane speaks first twice. Then they swap initiative again. Since the processes are autonomous, we need to stop them both with a quit program called jungle:quit().
Note about timeouts: The agents behave as if in a normal conversation. It is considered rude if one person always starts each conversation because the topic is set at the start. If agent-1 starts the converstion twice in a row then agent-1 should abdicate the starting position by changing the timesouts. An ethernet collision is another situation where timeouts must change to smooth communication. Actually when one agent always speaks first it is the opposite of a collsion, so we might call such a dialog situation an anticollision. If one agent always speaks first then it is more like a traditional client-server situation. To cure an anticollision, we must swap the relative magnitude of the timeouts.
Example program listing: jungle.erl
-module( jungle ). -compile(export_all). %% This program shows how chat-bot agents can exchange initiative(lead) while in conversation. %% Start with start(). %% End with quit(). start() -> register( tarzan, spawn( jungle, person, [ tarzan, 8000, "", jane ] ) ), register( jane, spawn( jungle, person, [ jane, 10000, "", tarzan ] ) ), "Dialog will start in 5ish seconds, stop program with jungle:quit().". quit() -> jane ! exit, tarzan ! exit. %% Args for person/4 %% Name: name of agent being created/called %% T: timeout to continue conversation %% Last: Last thing said %% Other: name of other agent in conversation person( Name, T, Last, Other ) -> receive "hi" -> respond( Name, Other, "hi there \n " ), person( Name, T, "", Other ); "slower" -> show( Name, "i was told to wait more " ++ integer_to_list(round(T*2/1000))), person( Name, T*2, "", Other ); "faster" -> NT = round( T/2 ), show( Name, "I was told to wait less " ++ integer_to_list(round(NT/1000))), person( Name, NT, "", Other ); exit -> exit(normal); _AnyWord -> otherwise_empty_the_queue, person( Name, T, Last, Other ) after T -> respond( Name, Other, "hi"), case Last of "hi" -> self() ! "slower", sleep( 2000), % give the other time to print Other ! "faster", person( Name, T, "", Other ); _AnyWord -> person( Name, T, "hi", Other ) end end. % respond( Name, Other, String ) -> show( Name, String ), Other ! String. % show( Name, String ) -> sleep(1000), io:format( " ~s -- ~s \n ", [ Name, String ] ). % sleep(T) -> receive after T -> done end. % ===========================================================>%
Sample output from: jungle.erl
Sample output: 18> c(jungle). {ok,jungle} 19> jungle:start(). jane_and_tarzan_will_start_in_5_seconds tarzan -- hi jane -- hi there tarzan -- hi jane -- hi there jane -- I was told to wait less: 5 tarzan -- I was told to wait more: 16 jane -- hi tarzan -- hi there jane -- hi tarzan -- hi there tarzan -- I was told to wait less: 8 jane -- I was told to wait more: 10 tarzan -- hi jane -- hi there tarzan -- hi jane -- hi there jane -- I was told to wait less: 5 tarzan -- I was told to wait more: 16 jane -- hi tarzan -- hi there 20> jungle:quit(). exit