Agent-Oriented Design

Agent-Oriented DesignTravis Steel

2

ObjectivesWhat is the Agent Paradigm?What is Agent-Oriented Design and how is it different

than OO?When to apply AOD techniques?When NOT to apply AOD techniques?

MAVs Lab, University of Texas at Dallas

3

OutlineDefinition of AgentMulti-Agent Systems (MAS)The Agent ParadigmAgent-Oriented DesignDomainsConclusion

MAVs Lab, University of Texas at Dallas

4

Agent:is an autonomous software entityhas objectives to satisfypossesses skills and can offer servicespossesses resources of its owncan communicate, cooperate, coordinate and negotiate

directly with other agentsacts in an environment that is partially perceived

MAVs Lab, University of Texas at Dallas

5

Example Structure of an Agent

MAVs Lab, University of Texas at Dallas

Virtual Agent

Interaction Module

Environment Perception Module

Agent Communication Module

Task Module

Task nTask 1 …

Knowledge Module

ExternalKnowledge Module

InternalKnowledge Module

Environment Model

Acquaintance Model

Self Model

Constraints Model

Planning & Control Module

Control

Execution

Planning

6

Multi-Agent System (MAS)A system that consists of a number of agents which interact with one another by exchanging knowledge and by negotiating with each other to achieve their own or some global goal

Common Characteristics:non-deterministicdistributedadaptive

MAVs Lab, University of Texas at Dallas

7

The Agent ParadigmProgramming & Design paradigms:

machine code -> prog. languages -> sub-routines -> abstract data types -> objects -> agents

Main differences between OOD and AOD:

Concept of “Agent” is not limited to a certain application or domain

MAVs Lab, University of Texas at Dallas

Object-Oriented Agent-Oriented

Objects Agents

Nonautonomous Autonomous

Interactions are pre-defined Constraints are pre-defined

Satisfy Requirements Satisfying Goals (thus Req’s)

Outcome is usually predictable Outcome is usually unpredictable

8

Agent-Oriented Design

MAVs Lab, University of Texas at Dallas

Agents are the primary design abstractionDefine roles, constraints, and goalsDefine and coordinate agent rolesAssign agent goals that help achieve the system

objectives

9MAVs Lab, University of Texas at Dallas

Agent Goal Diagram

cite [1]

10

Agent-Oriented DesignAOD is new, so there are many methodologies, and few

standardsMethodologies and Languages:

Agent-UML (FIPA [3] and OMG): extension of UMLAgent Communication Language (ACL)Knowledge Query Meta Language (KQML)

MAVs Lab, University of Texas at Dallas

11

Example Patterns forAgent Interaction

MAVs Lab, University of Texas at Dallascite [2]

Bidding Matchmaker

12

Other Differences BetweenOOD and AODObject states vs. Agent rolesConcurrent & context-based interactions in sequence

diagramsRole changes in collaboration diagramsActivity diagrams show how roles effect high-level

interaction in various situationsPackages can classify agents based on roles or goalsDeployment diagrams describe how agents are deployed

in their host environment

MAVs Lab, University of Texas at Dallas

13

Domains for MASTransportationMilitaryMedicineWeb SearchFinanceSocial Networking

MAVs Lab, University of Texas at Dallas

14

ConclusionAOD is a strong approach to solving problems that are

considered too complex using other approaches.The Agent Paradigm provides an alternate way to

conceptualize and design software systems.The Agent Paradigm is not limited to a certain application

or domain but can be applied in a variety of circumstances.

AOD can simplify complex problems and over-complicate simple problems.

MAVs Lab, University of Texas at Dallas

15

Questions?

MAVs Lab, University of Texas at Dallas

16

References1. Discussing strategies for software architecting and

designing from an Agent-oriented point of view Anna Perini, Angelo Susi

2. Introspecting Agent-Oriented Design Patterns Manuel Kolp, T. Tung Do, Stéphane Faulkner and T. T.

Hang Hoang3. The Foundation for Intelligent Physical Agents:

http://www.fipa.org/

MAVs Lab, University of Texas at Dallas