1 / 76

DALI Multiagent System

DALI Multiagent System. Handbook and Examples. How to create and use a DALI Agent: Summary. How the DALI Interpreter starts (provided you have installed Sicstus Prolog) How to create an agent: the DALI initialization file How to write a DALI logic program

gittel
Download Presentation

DALI Multiagent System

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. DALI Multiagent System Handbook and Examples

  2. How to create and use a DALI Agent: Summary • How the DALI Interpreter starts • (provided you have installed Sicstus Prolog) • How to create an agent: • the DALI initialization file • How to write a DALI logic program • Architecture of the DALI Interpreter • Examples

  3. How the DALI Interpreter starts active_dali.pl/ active_dali.exe active_server.pl/ active_server.exe . . . active_user.pl/ active_user.exe active_dali.pl/ active_dali.exe In order to start the DALI Multiagent System it’s necessary to activate the server, next the user module, finally one or several DALI agents.

  4. file_dali.txt active_server.pl active_user.pl active_dali.pl active_server.pl/ active_server.exe How the Server starts bin demo program We can activate the DALI server either by invoking the the executable file ‘active_server.exe’ or, via the Sicstus Prolog shell, by loading the file ‘active_server.pl’. This is the command to load the server, that you can fond in the demo directory: load_files('C:/Programmi/SICStus Prolog 3.11.1/bin/demo/active_server.pl').

  5. active_user.pl/ active_user.exe How the User Module starts We can activate the DALI user module either by using the ‘active_user.exe’ executable file or, via the Sicstus Prolog shell, by loading the file ‘active_user.pl’. This is the command to load the file that you can find in the demo directory: load_files('C:/Programmi/SICStus Prolog 3.11.1/bin/demo/active_user.pl'). The user module opens a user window for communicating with DALI agents, as explained later on. The name of the Receiver agent The language of the message The list of terms useful to interprete a message The name of the Sender agent The content of the message

  6. Initialization file active_server.pl active_user.pl active_dali.pl help.txt help.txt bin demo program active_dali.pl/ active_dali.exe How the DALI Interpreter starts We can activate the DALI Interpreter either by using the executable file ‘acive_dali.exe’ or, via the Sicstus Prolog shell, by loading the file ‘active_dali.pl’. This is the command to load the file that you can find in the demo directory: load_files('C:/Programmi/SICStus Prolog 3.11.1/bin/demo/active_dali.pl'). The interpreter will ask the name of the initialization file: Insert the path and the name of the initialization file: |: 'demo/help.txt'. And, after this information, the agent will be activated: .................. Actived Agent ................... DALI logic program In this way it is possible to activate several DALI agents having different names and logic programs.

  7. DALI initialization file The initialization file <init>.txt must be in the directory “demo”, and contains the following informations: • The name of the file that contains the DALI logic program; • The name of the agent; • The adopted ontology (a txt file); • The adopted language (Italian,English,…) in the communication acts; • The name of the file containing the tell/told communication constraints; • The name of the communication library; • The agent’s abilities ( kind of work, hobbies,…) These parameters are grouped in a string with prefix ‘agent’, according to the above order, with the syntax exemplified below.

  8. DALI initialization file: example Example: content of the initialization file agent('demo/program/italian',gino, 'demo/pippo_ontology.txt',italian, ['demo/communication'], ['demo/communication_fipa'],[tourist]). The path is specified starting either from the directory bin if we use the Sicstus Prolog shell or starting from the directory where the interpreter is if we use the DALI executable file.

  9. DALI initialization file: example Precisely: ‘gino’ is the name of the agent, ‘ontology.txt’ is the ontology file in the directory ‘demo’ and Italian is the language spoken by the agent. Finally, ‘communication’ is the file ‘communication.con’ in the directory ‘demo’ containing the tell/told constraints and ‘communication_fipa’ is the library with the fipa communication’s primitives. The last papameter suggests that the agent is a tourist. At this point, we can write a DALI logic program with .txt extension and put it in the directory program.

  10. file_dali.txt ontology.txt communication_fipa.txt communication.con active_dali.pl Generate a DALI initialization file In the initialization string (see above) we put the initialization file under the directory demo and the DALI logic program file ‘file_dali.txt’ in the directory program: bin demo program

  11. active_user.pl/ active_user.exe How the User Module works The user module allows the user to communicate with an existing DALI agents, by means of the DALI primitive send_message(Content,Sender) where Sender=user and Content is an external event we want the agent to perceive. will ask the following arguments: Insert name of addressee |: pippo. Insert Language |: italian. Insert Ontology |: []. Insert From |: user. Insert message |: send_message(danger,user). The name of the Receiver agent The language of the message The list of terms useful to interprete a message The name of the Sender agent The content of the message

  12. How to use Ontologies in the User Module (simple way, to be further developed) If we have for instance, in the DALI logic program, the following reactive rule: dangerE:>once(ask_for_help). ask_for_help:-call_policeA. call_police:<have_a_phoneP. ask_for_help:-screamA. If we tell the agent that ‘pericolo’ is equivalent to ‘danger’, we can then send the message send_message(pericolo,user). Insert name of addressee |: pippo. Insert Language |: italian. Insert Ontology |: [danger]. % this is to be replaced in place of the following event Insert From |: user. Insert message |: send_message(pericolo,user).

  13. Write DALI logic program A Dali program is a file txt whose content is a DALI program. Now we shortly recall the main features of this language by writing a DALI agent: External events: The external events are syntactically indicated by the postfix E. When an event enters the agent from its external world, she can perceive it and decide to react. The reaction is defined by a reactive rule which has in its head that external event. The special token :>, used instead of :-, indicates that the reactive rule performs forward reasoning. If we write in the txt file this simple reactive rule: alarm_clock_ringsE:>stand_upA the agent observes the following behavior. We use the user module to send to the agent the external event alarm_clock_rings.

  14. .................. Actived Agent ................... make(stand_up) Write DALI logic program:External Event Insert name of addressee |: pippo. Insert Language |: italian. Insert Ontology |: []. Insert From |: user. Insert message |: send_message(alarm_clock_rings,user). Nome file: alarm.txt The DALI agent will do the action contained in the body of the reactive rule.

  15. Write DALI logic program:Internal Event Internal events: The internal events define a kind of individuality of a DALI agent, making her proactive independently of the environment, of the user and of the other agents, and allowing her to manipulate and revise her knowledge. An internal event is syntactically indicated by the postfix I, and its description is composed of two rules. The first one contains the conditions (knowledge, past events, procedures, etc.) that must be true so that the reaction (in the second rule) may happen. If we write in the txt file those two rules: i_am_lazy:-alarm_clock_ringsP not(stand_upP) i_am_lazyI:> i_take_a_vacation_dayA. the agent exhibits the following behavior. We use the user module to send to the agent the past event alarm_clock_ringsP but this past event could be the past form of the external event as specified in the previous paragraph, and arrived before.

  16. .................. Actived Agent ................... make(i_take_a_vacation_day) Write DALI logic program:External Event Insert name of addressee |: pippo. Insert Language |: italian. Insert Ontology |: []. Insert From |: user. Insert message |: confirm(alarm_clock_rings,user). Nome file: alarm_clock.txt The DALI agent will make the action contained in the reaction of the internal event.

  17. Write DALI logic program:Present Event Present events: When an agent perceives an event from the external world it does not necessarily react to it immediately: she has the possibility of reasoning on the event, before (or instead of) triggering a reaction. Reasoning also allows a proactive behavior. In this situation, the event is called present event and is indicated by the suffix N. arrives_someone:-bell_ringsN. arrives_someoneI:>get_dressedA. get_dressed:< get_undressedP. bell_ringsE:>open_the_doorA. In this case, when we send the external event bell_rings to the agent, she makes the actions get_dressed and open_the_door if the internal event has been processed before the external event. Else, she does only the action open_the_door. In small DALI programs is unlikely to observe the reaction to the internal event, because the processing of the external events is faster and the interpreter, after the action open_the_door, erases the reaction to the internal event.

  18. .................. Actived Agent ................... make(open_the_door) make(open_the_door) make(open_the_door) make(open_the_door) make(get_dressed) make(open_the_door) make(open_the_door) make(open_the_door) make(open_the_door) Write DALI logic program:Present Events Insert name of addressee |: pippo. Insert Language |: italian. Insert Ontology |: []. Insert From |: user. Insert message |: send_message(bell_rings,user). Nome file: wear.txt

  19. Write DALI logic program:Actions Actions: • Simple actions: The action in the DALI program is specified as actionA. For example i_go_to_bedA, i_take_the_busA, … When the agent does an action, on the prolog shell you can observe make(action). • Messages as actions: From a DALI program, a message can be sent by writing: messageA(To, Content) where To is the name of the agent that must receive the communication act and Content is the DALI/FIPA primitive.

  20. Write DALI logic program:Actions We now describe the primitives used in a DALI program. Sender is the agent that sends the message. • send_message(External_event, Sender) this primitive is used to call an external event within a reactive rule; • confirm(Assertion, Sender) this primitive is used to assert a fact within a DALI agent; • disconfirm(Assertion, Sender) this primitive is used to retract a fact within a DALI agent; • propose(Action, List_Conditions,Sender) this primitive is used to propose an agent to do an action contained within a DALI program. The agent can respond by accepting to do the action (accept_proposal) or by rejecting the proposal (reject_proposal) if the conditions in the message are false. When the agent decides to make the action, she verifies if the conditions of the action rule are true. In this case, the agent does the action or else she sends to the other agent a failure message.

  21. Write DALI logic program:Actions • execute_proc(Head, Sender) this primitive is used to invoke the head of a generic rule (procedure) in the DALI program; • query_ref(Fact, N, Sender) this primitive is used to answer an agent on some information about a not ground Fact. N is the number of the requested matchings; • agree(Fact, Sender) this primitive is used to answer an agent knowing a ground Fact; • cancel(Action, Sender) this primitive is used to communicate to an agent to cancel a requested action. Also in this case, it is difficult to see the effect of this primitive because the agent does immediately the action. To see the effect, the queue of actions must contain several items.

  22. Write DALI logic program: Goals A goal is an objective that an agent must reach. In the DALI language, a goal is a particular internal event that the interpreter begins to attempt when it is invoked. The goal has a postfix G. How it works: Environment State ...,goal1G,... goal1:- condition11,...,condition1k subgoal11G,...,subgoal1NG,subgoalP11,...,subgoalP1N. ... goal1:-conditionm1,...,conditionmk subgoalGm1,...,subgoalGmN,subgoalP11,...,subgoalP1N. goal1I:>action1,...,goal2G,...,actionk. PLANNER Goal Plan to reach a goal Actions

  23. Write DALI logic program: Goals We consider a simple example: an agent must wear socks and shoes. goE:>put_shoesG. put_shoes:-put_right_shoeG,put_left_shoeG,right_shoe_onP, left_shoe_onP. put_shoesI:>left_and_right_shoes_onA, retractall(past(_,_,_)). put_right_shoe:-put_right_sockG,right_sock_onP. put_right_shoeI:>right_shoe_onA. put_left_shoe:-put_left_sockG,left_sock_onP. put_left_shoeI:>left_shoe_onA. put_right_sock:-have_right_sockP. put_right_sockI:>right_sock_onA. put_left_sock:-have_left_sockP. put_left_sockI:>left_sock_onA.

  24. .................. Actived Agent ................... make(right_sock_on) make(right_shoe_on) make(left_sock_on) make(left_shoe_on) make(left_and_right_shoes_on) Write DALI logic program: Goals Insert name of addressee |: pippo. Insert Language |: italian. Insert Ontology |: []. Insert From |: user. Insert message |: confirm(have_left_sock,user). Insert name of addressee |: pippo. Insert Language |: italian. Insert Ontology |: []. Insert From |: user. Insert message |:confirm(have_right_sock,user). Insert name of addressee |: pippo. Insert Language |: italian. Insert Ontology |: []. Insert From |: user. Insert message |: send_message(go,user). Nome file: shoes.txt

  25. Write DALI logic program: Past events • Past events: • A past event, indicated by the suffix P, is: • an external or internal event after a reaction; • an executed action; • a reached goal or subgoal; • a fact communicated using a confirm primitive. Insert name of addressee |: pippo. Insert Language |: italian. Insert Ontology |: []. Insert From |: user. Insert message |: confirm(rain,user). past(rain,timestamp, user) is recorded within DALI shell. This past event is called using the string with postfix P. For example, rainP.

  26. The architecture of a DALI agent Pre-processing file.txt Processing Events Actions Goals Communication module Communication module

  27. file.ple file.txt file.plf file.pl The DALI interpreter generates some auxiliary files for each agent. These files are put in the same directory as the txt DALI file. The files that the interpreter generates from DALI txt file

  28. A simple txt DALI file We use this txt file to show how the interpreter works creating ple, plf andpl files. In order to show the ‘naming’ process we use a new example containing variables. External event Reaction rule dangerE:>once(ask_for_help). ask_for_help:-call_policeA. call_police:<have_a_phoneP. ask_for_help:-screamA. remain_at_home:-dangerP,call_policeP. remain_at_homeI:>go_to_bathroomA, close_the_doorA. go_out:-dangerP,screamP. go_outI:>go_to_neighbourA. Past event Action rule Action Internal event

  29. What the Interpreter records in the ple file External events [danger]. [remain_at_home,go_out,external_refused_action_propose(A,Ag), refused_message(AgM,Con)]. [call_police,scream,go_to_bathroom,close_the_door,go_to_neighbour,message(Ag,inform(query_ref(X,N),values(L),A)),message(Ag,refuse(query_ref(variable),motivation(refused_variables),A)),message(Ag,inform(query_ref(X,N),motivation(no_values),A)),message(Ag,inform(agree(X),values(yes),A)),message(Ag,inform(agree(X),values(no),A))…]. [call_police]. []. []. []. Internal events Conditions Actions Present events Goals to reach This file is used by the interpreter to manage the behavior of the agent through the classes of the events, the actions, goals,… Goals to test

  30. The ‘naming’ of variables in the pl file Dali pl program with reified variables eve(predator_attacks(var_X)):- once(try_to_escape(var_X)).try_to_escape(var_X):- a(fly(var_X)).try_to_escape(var_X):- a(run(var_X)).cd(fly(var_X)):- evp(bird(var_X)), not(evp(abnormal(var_X))). Interpreter work Dali txt program predator_attacksE(X):> once(try_to_escape(X)). try_to_escape(X):-flyA(X). try_to_escape(X):-runA(X). fly(X):<birdP(X),not(abnormalP(X)). The interpreter generates a pl file where all rules are subjected to ‘naming’ process. All variables has been transformed to costants using the suffix ‘var_’.

  31. The directives of plf file We use this file to set some parameters that determine the behavior of the agent relatively to external, internal, past events and actions . Past event We can decide how long or until which condition a past event must be kept in the memory of an agent: past_event(Event,Seconds). (the past event is kept in memory some Seconds) past_event(Event,forever). (the past event is kept in memory forever) past_event(Event,until(Cond)). (the past event is kept in memory until Condition) External event We can decide if an external event must be processed with a normal or high priority: external_event(Event,normal). external_event(Event,high). Action We can decide if an action must be processed with a normal or high priority: action(Action,normal). action(Action,high). Action/Message We can decide if to submit a message to tell the tell check: mod(Action, check).

  32. The directives of plf file Now we examine the directives on internal events. • Internal event • We can set several parameters for an internal event in order to tune the • Interpreter behaviour: • the frequency(seconds) with which the interpreter attempts the internal event; • how many times the agent must react if the internal event is true (1,2,..,forever); • when the agent must react: this parameter is ‘true’ if the reaction must happen • forever or else we can link the reaction to some past events belonging to body • of the first rule of the internal event; when the past event inside a ‘change list’ • is modified the agent will be able to react again. • For example, if we have the internal event: • think:-rainP, go_outP,buy_umbrellaP. • thinkI:>open_the_umbrellaA. • we can set parameters as: • internal_event(think,3,1,change([rain,go_out]),forever). • when/until when the interpreter must attempt the internal event: • until_date(Date): till certain date; • until_cond(Condition): when the Condition is false; • forever: forever

  33. Some directives of plf example file action(call_police,normal). action(scream,normal). action(go_to_bathroom,normal). action(close_the_door,normal). action(go_to_neighbour,normal). external_event(danger,normal). past_event(danger,20). past_event(remain_at_home,20). past_event(go_out,20). past_event(call_police,20). past_event(scream,20). past_event(go_to_bathroom,20). past_event(close_the_door,20). past_event(go_to_neighbour,20). internal_event(remain_at_home,3,forever,true,until_cond(past(remain_at_home))). internal_event(go_out,3,forever,true,until_cond(past(go_out))). mod(message(_,inform(_,motivation(refused_message),_)),check). The action is put in the queue with normal priority This past event is kept in memory 20 seconds This internal event is attempted every 3 seconds The agent will react forever We specify no conditions This internal event is attempted until the past event remain_at_home becomes true This message is submitted to tell check

  34. The Communication architecture Incoming message The message passes this level only if the corresponding told rule is true. TOLD CHECK If the agent doesn’t know the content of the message, she calls the meta-level and uses the ontology and/or other properties in order to understand the communication act. META LEVEL DALI INTERNAL INTERPRETER The message, submitted to tell check, is sent only if the corresponding tell rule is true. TELL CHECK Outcoming message

  35. Incoming message TOLD CHECK The Told Check level Each DALI agent has a ‘con’ file, specified in the initialization file, that contains the told/tell rules. These rules, external to interpreter, can be modified by the user. The structure of a told rule is: told(Sender,Content):-constraint1,…,constraintn. In this example, an agent receives a message, using the primitive send_message, only if the Sender agent isn’t an enemy: told(Ag,send_message(_)):-not(enemyP(Ag)). A message that does not go through the told level is eliminated. The agent Sender receives an inform message asserted as a past event.

  36. The Meta-Level META LEVEL Each DALI agent uses an (optional) meta procedure written in the ‘communication.con’ file that specifies how the entity can interprete an unknown message. This procedure can be modified by the user. In the initialization file the user can specify the ontology txt file that the agent must use. This file contains some strings with the following prefix and parameters: ontology(Agent_name, Term1,Term2). where Agent_name is the name of the agent that can use the association of terms Term1 and Term2. In this file the user can write some properties of terms useful to meta-procedure. For example: symmetric(love(julie,tom)). specifies that the term love is symmetric.

  37. The Tell Check level TELL CHECK Outcoming message The user can set the mod string in the plf file in order to submit a message to tell check. In this case, the message is actually sent only if the corresponding tell rule is true. The tell rules can be modified by the user. The structure of a tell rule is: tell(Receiver,Sender,Content):-constraint1,…,constraintn. In this example, an agent sends a message, using the primitive send_message, only if she has a trust greater than 4 in the Receiver agent: tell(To,_,send_message(_)):-trustP(_,To,N),N>4. A message that does not go through the tell level is eliminated.

  38. Examples – A dangerous situation… One Agent!

  39. help.txt active_server.pl active_user.pl active_dali.pl help.txt Examples – A dangerous situation… Initialization file name: help.txt DALI logic file name: help.txt bin demo program

  40. Examples – A dangerous situation… DALI logic program help.txt dangerE:>once(ask_for_help). ask_for_help:-call_policeA. call_police:<have_a_phoneP. ask_for_help:-screamA. remain_at_home:-dangerP,call_policeP. remain_at_homeI:>go_to_bathroomA, close_the_doorA. go_out:-dangerP,screamP. go_outI:>go_to_neighbourA.

  41. Examples – A dangerous situation… Run the DALI logic program Inserire il percorso ed il nome del file di inizializzazione: |: 'demo/help.txt'. Insert name of addressee |: pippo. Insert Language |: italian. Insert Ontology |: []. Insert From |: user. Insert message |: send_message(danger,user). .................. Actived Agent ................... make(scream) make(go_to_neighbour)

  42. Examples – A dangerous situation… If the agent has a phone… New message Insert name of addressee |: pippo. Insert Language |: italian. Insert Ontology |: []. Insert From |: user. Insert message |: confirm(have_a_phone,user). New message Insert name of addressee |: pippo. Insert Language |: italian. Insert Ontology |: []. Insert From |: user. Insert message |: send_message(danger,user). she reacts differently! .................. Actived Agent ................... make(call_police) make(go_to_bathroom) make(close_the_door)

  43. Examples – Error Recovery planning… Two Agents!

  44. recovery.txt active_server.pl active_user.pl active_dali.pl recovery.txt sensor.txt sensor.txt Examples – Error Recovery planning… Initialization file name: recovery.txt and sensor.txt DALI logic file name: recovery.txt and sensor.txt bin demo program

  45. Examples – Error Recovery planning… DALI logic program recovery.txt error_recovery(M,M1):-go_rightP(_,M),goal(M),informP(reality(M,M1),sensor),M\=M1. error_recovery(M,M1):-go_leftP(_,M),goal(M),informP(reality(M,M1),sensor),M\=M1. error_recovery(M,M1):-go_forwardP(_,M),goal(M),informP(reality(M,M1),sensor),M\=M1. error_recoveryI(M,M1):>recovery_errorA(M,M1),drop_pastA(reality(M,M1)). reached_goal(M):-go_rightP(_,M),goal(M). reached_goal(M):-go_leftP(_,M),goal(M). reached_goal(M):-go_forwardP(_,M),goal(M). reached_goalI(M):>clause(agent(Ag),_), messageA(sensor,send_message(what_about_my_position(M,Ag),Ag)). … right(exit1,bank1). right(exit2,bank2). left(bank1,hospital1). forward(bank2,bank1). goal(hospital1).

  46. Examples – Error Recovery planning… DALI logic program sensor.txt what_about_my_positionE(M,Ag):>once(examine_position(M,Ag,_)). examine_position(M,Ag,M1):-messageA(Ag,inform(reality(M,M1),sensor)),wrong_positionA(M,Ag). message(_,inform(reality(M,M1),sensor)):<errorP(M,M1). examine_position(M,Ag,_):-right_positionA(M,Ag).

  47. Error Recovery planning The environment… Hospital Bank2 Bank1 Exit1 Exit2 STATION

  48. Error Recovery planning First situation: the agent exits from Exit1 and goes to Hospital1 Hospital Bank2 Bank1 Exit1 make(go_right(exit1,bank1)) make(go_left(bank1,hospital1)) Exit2 confirm(i_am_at_exit(exit1),user) STATION

  49. Insert the path and the name of the initialization file: |: 'demo/sensor.txt'. Insert the path and the name of the initialization file: |: 'demo/recovery.txt'. Examples – Error Recovery planning… Run the DALI logic programs recovery.txt and sensor.txt Insert name of addressee |: pippo. Insert Language |: italian. Insert Ontology |: []. Insert From |: user. Insert message |: confirm(i_am_at_exit(exit1),user). sensor .................. Actived Agent ................... make(right_position(hospital1,pippo)) robot pippo .................. Actived Agent ................... make(go_right(exit1,bank1)) make(go_left(bank1,hospital1)) send_message_to(sensor,send_message(what_about_my_position(hospital1,pippo), pippo),italian,[])

  50. Error Recovery planning Second situation: the agent exits from Exit2 and goes to Hospital1 Hospital Bank2 Bank1 make(go_right(exit2,bank2)) make(go_forward(bank2,bank1)) make(go_left(bank1,hospital1)) Exit1 Exit2 confirm(i_am_at_exit(exit2),user) STATION

More Related