Using Environment Properties to Select Agent Architectures

1
Using Environment Properties to Select Agent
Architectures

• Carla Silva, Jaelson Castro, Patrícia Tedesco

ctlls, jbc, pcart_at_cin.ufpe.br
Laboratório de Engenharia de Requisitos -
www.cin.ufpe.br/ler
2
Agenda

• Motivation
• NFR-Framework
• Agent Types
• Environment Types
• Proposal
• Case Study
• Contributions
• Future Work

3
Motivation

• There are different types of agents
• Agents inhabit different types of environments
• Some environments are more complex than others
• How well an agent can behave depends on the
  nature of the environment

4
Motivation

• Using environments characteristics to
  systematically guide selection among agent
  architecture alternatives.

5
Solution

• We adapted the NFR-framework
• treats Non-Functional Requirements as goals
• allows AND-decomposition and OR-decomposition of
  goals
• evaluates the influence of a design decision to
  the satisfaction of a goal

Chung, Nixon, Yu and Mylopoulos 2000
6
NFR-Framework
Softgoal Interdependency Graph
7
NFR-Framework
Propagation Mechanism
8
Proposed Approach

• Using the NFR-framework to evaluate the
  suitability of certain agent architectures with
  respect to certain environment types.

9
Agent Types

• Reactive Agents
• Deliberative Agents
• Purely Deliberative
• Learning Agents
• Hybrid Agents

WEISS 2000, RUSSEL and NORVING 2002
10
Environment Types

• Accessible vs. inaccessible
• Deterministic vs. non-deterministic
• Episodic vs. non-episodic
• Static vs. dynamic
• Discrete vs. continuous

WEISS 2000, RUSSEL and NORVING 2002
11
Softgoal Interdependency Graph
12
Softgoal Interdependency Graph
13
Correlation Catalogue
14
Case Study

• A multi-agent system for an Electronic Newspaper,
  called e-News
• allows users to obtain news from a newspaper web
  site.
• news related to a specific subject are extracted
  from different and distributed news agencies in
  the Internet, translated and merged in order to
  provide quality news.

15
Case Study

• Step1. Characterising the system environment

inaccessible, non-deterministic, non-episodic
dynamic and continuous

• Step2. Matching the proper agent architecture

16
Correlation Catalogue
17
Case Study

• Further decide if these agents are purely
  deliberative, hybrid or learning.
• This decision depends on the application domain
  and designer experience.

• The agents for the e-News system use a purely
  deliberative architecture
• they do not need to perform any reactive or
  learning behaviour.

18
Contributions

• Our approach aims to promote an efficient
  utilization of agent architectures in order to
  achieve a successful agent performance.
• The Correlation Catalogue can be used by any
  agent oriented methodology.

19
Future Work

• Propose a systematic method to select among
  hybrid, learning and purely deliberative
  architectures
• Involve other types of agents
• Mobile
• Etc (TROPOS)

20
Tropos _at_ SBES05

• Describing Agent-Oriented Design Patterns in
  Tropos
• Carla Silva, Jaelson Castro, Patrícia Tedesco and
  Ismênia Silva
• Support for Requirement Traceability The Tropos
  Case
• Rosa Candida Pinto, Carla Silva and Jaelson
  Castro
• ST1 Requisitos
• Quarta-feira às 11.00hs
• Sala B2B3