Extending the AAII methodology for programming agents in 3APL

1
Extending the AAII methodology for programming
agents in 3APL

• Pierre-Yves Schobbens
• Jean-Christophe Trigaux
• François Vermaut

2

• Aim Provide a method to translate AAII models
  into 3APL agent specifications.
• Plan
• 1) Introduce the AAII methodology and the 3APL
  language
• 2) Transform AAII model into 3APL structures
• 3) Interaction model for the AAII methodology

3
AAII - Methodology

• External viewpoint
• Agent model
• Interaction model
• Internal viewpoint
• Belief model
• Goal model
• Plan model

4
AAII Agent model

• Based on UMLs class diagram notation
• Types instances
• Belief-state-set goal-state-set
• Initial-belief-set initial-goal-set

5
AAII Interaction model

• No standard notation
• ? users choice
• MSC HMSC notation from MAS-CommonKADS

6
AAII Belief model
7
AAII Goal model

• Expressed with predicates of the FOL with modal
  operators (achieve(X), test(X), verify(X))
• Directly linked to a belief
• 3 Types
• Achievement goals
• Test goals
• Verify goals

8
AAII Plan model

• Based on UMLs statechart notation
• Plan nodes transitions
• 3 types of nodes
• Start nodes
• Internal nodes
• End nodes (success, fail)
• Transition triggered by an event, if a certain
  condition is satisfied

9
AAII Plan model example
10
3APL - language

• Koen Hindriks Steven Anker
• 3APL  An Abstract Agent
  Programming Language 
• Based on BDI theory
• Originalities
• Combine logic and imperative programming
• Pratical reasoning rules (PR-rules)

11
3APL-Agent
12
3APL Agent Specifications

• Belief
• ? Formula of the FOL
• Basic Actions
• ? ? ? ?
• Goals
• Bact ? Goal
• Atom ? Goal
• If ? ? Beliefs, then ?? ? Goal
• If ?1, ?2 ? Goal, then (?1?2), (?1?2),
  (?1?2) ? Goal
• PR-rules
• head ? guard body

13
Beliefs goals derivation (1)

• Beliefs Derivation
• AAII Initial-belief-set ? 3APL beliefBase
• Syntactic equivalence (predicate of the FOL)
• Semantic difference
• AAII beliefs use trivalent logic ltgt 3APL beliefs
  use boolean logic
• Solution
• Limit AAII beliefs to boolean logic
• Goals Derivation
• AAII achievement or test goals ? 3APL achievement
  or test goals
• Syntactic differences
• Modal operators in AAII goals
• Only ? in 3APL test goals

14
Beliefs goals derivation (2)

• Semantic differences
• AAII goals make belief true ltgt 3APL goals
• AAII achievement goals trigger a plan if
  associated belief is false
• ltgt 3APL goals PR-rules
• Solutions
• In 3APL, explicitly assert corresponding belief
  of a goal when the activated plan succeeds.
• In 3APL, abort the execution of a PR-rule if the
  associated belief is true.

15
Plan derivation(1)

• AAII plans are based on UMLs statechart notation
• AAII plans transitions ? 3APL PR-rule
• AAII plans events (goals) ? 3APL goals in the
  head of
• the PR-rule
• AAII plans actions ? 3APL goals in the body
  of the PR-rule

16
Plan derivation(2)

• 3APL PR-rules store less information than AAII
  plans
• AAII plans are splitted into several PR-rules

17
Interaction modeling

• No standard notation
• MSC ? model simple communication between agents
• HMSC ? model higher level protocols

18
Interaction model derivation(1)

• Messages from MSC diagrams ? 3APL beliefs
• Message events (received message) ? Waiting for
  the corresponding belief to become true
• Sending a message ? Asserting a belief in
  receivers beliefBase

19
Interaction model derivation(2)
20
Contributions

• Implement 3APL agents at a higher level of
  abstraction
• Extend AAII methodology with MSC and HMSC as
  interaction model
• Bridge the gap between agent systems modeling and
  agent systems programming.
• Study case in the e-Business domain

Future works

• Develop a tool based on this method