Imperative and Declarative Programming Languages for MultiAgent Systems

1
(Imperative and) Declarative Programming
Languages for Multi-Agent Systems

• João Alexandre Leite

CENTRIA - New University of Lisbon http//centria
.di.fct.unl.pt/jleite/
Ljubljana, March 1st, 2005
PROMAS_at_AL3-TF2
2
Declarative vs Imperative

• Declarative description
• A grandchild to x is a child of one of xs
  children.
• Imperative description I
• To find a grandchild of x, first find a child of
  x. Then find a child of that child.
• Imperative description II
• To find a grandchild of x, first find a
  parent-child pair and then check if the parent is
  a child of x.
• Imperative description III
• To find a grandchild of x, compute the factorial
  of 123, then find a child of x. Then find a child
  of that child.

• Imperative implementation
• read(person)
• for i 1 to maxp do
• if pi1 person then
• for j 1 to maxp do
• if pj1 pi2 then
• write(pj2)
• fi
• od
• fi
• od
• Declarative implementation
• gc(X,Z) - c(X,Y), c(Y,Z).

3
Declarative vs Imperative
4
Logic Based MAS

• ConGolog De Giacomo et al.
• 3APL Dastani et al.
• IMPACT Subrahmanian et al.
• Conc. MetateM Fisher et al.
• AgentSpeak(L) Rao et al.
• DyLog Baldoni et al.
• ?hhf Delzanno et al.
• AGENT 0 Shoham

• FLUX Thielscher et al.
• JASON Bordini et al.
• DALI Costantini et al.
• KARO van Linder et al.
• PROSOCS Stathis et al.
• MINERVA Leite et al.
• ALIAS Ciampolini et al.
• ...

5
CLIMA DALT

• Constantly increasing number of submitted papers
  (35 papers each in last editions)
• Covering most topics seen in current MAS
  research

• Negotiaton
• Planning
• Protocols
• Learning
• Adaptation
• Cooperation
• Communication
• Verification

• Reasoning
• Institutions
• Social Norms
• Architectures and Environments
• Trust
• Belief Revision and Update
• Programming Languages
• Software Engineering

6
Implementations
How are these systems implemented?

• PROLOG
• ConGolog
• Conc. MetateM
• DyLog
• ?hhf
• FLUX
• DALI
• MINERVA
• ALIAS

• Common LISP
• AGENT 0

• JAVA
• IMPACT
• JASON

• JAVA PROLOG
• 3APL
• PROSOCS

7
Some thoughts...

• On the opportunities MAS provides to DLs...
• The high (and increasing) complexity of MAS calls
  for DLs
• Declarative Languages are guilty until proven
  innocent!!
• On speed of DLs...
• Drop the its only for prototyping excuse
• Tackle proper implementations
• Work together with major DL implementation groups
  (e.g. XSB...)
• On ease of programming DLs and training costs...
• Train students at undergraduate level

8
Proposed Questions...

• Generic DLs (not developed ad-hoc for MAS
  development) such as PROLOG, Oz, Haskell, Scheme,
  ... could be used to develop such systems, but is
  it really more helpful?
• Pure declarative implementations or hybrid?
• How can DLs be used together with imperative
  ones?
• How can DLs facilitate the MAS development?
• It seems that DLs can be applied to define the
  high-level behavior of an agent. Where else?