Agent Oriented Programming

1
Agent Oriented Programming

• EEL 5937 Multi Agent Systems
• Lecture 26, April 10, 2003
• Lotzi Bölöni

2
Software development methodologies

• Progress in software engineering has been made
  through new, powerful abstractions
• Procedural abstraction
• Abstract data types
• Objects
• Components
• If agents are to realize their potential as a
  software engineering paradigm, it is necessary to
  develop techniques specifically tailored to them.
• Current approaches (e.g. rational unified
  process) fails to capture the
• Agents flexible, autonomous behaviour
• The richness of interactions
• Complexity of the agents organizational
  structures
• Mobility and mutation

3
Why not object oriented techniques?

• While there are superficial similarities between
  agents and objects
• But the granularity is to fine grained, operating
  at the wrong level of abstraction.
• An agent so represented may appear strange maybe
  a single public function receive which
  encapsulates everything.
• Secondary models such as use cases and
  interaction diagrams can be usefully used.
• But the object model and the class diagram does
  not help us with agents.
• Because the dependencies between agents are not
  static.

4
Agent methodologies

• Numerous methodologies were proposed, which can
  be categorized as follows
• Techniques using OO modelling techniques as their
  basis (GAIA, Agent UML, the BDI technique
  proposed by Kinny et. al.)
• Methodologies based on techniques form knowledge
  engineering, providing formal, compositional
  modelling languages suitable for verification of
  system structure and function.
• Methodologies taking existing formal languages
  such as Z, and specify agents systems in this
  framework.
• Methodologies developed specifically for agent
  systems. E.g. Cassiopeia based on the Contract
  Net model.

5
GAIA
6
What is it?

• A software methodology targeted to the analysis
  and design of agent-based systems.
• It is intended to allow an analyst to go
  systematically from a statement of requirements
  to a design that is sufficiently detailed to be
  implemented directly.
• It is loosely based on object oriented
  technologies, such as FUSION.
• Analysis and design can be seen as a process of
  developing increasingly detailed models of the
  system to be constructed.

7
Models of the Gaia methodology
8
What kind of agents are handled

• Coarse-grain agent systems. E.g. each agent
  Unix process.
• The goal is to obtain a system that maximizes
  some quality measure, but which might be
  sub-optimal for individual agents. Gaia is not
  intended for systems that admit the possibility
  of true conflict.
• Agents can be heterogeneous.
• The organization structure of the system is
  static, inter-agent relationships do not change
  over time.
• The abilities of the agents and their services do
  not change over time.
• The system contains a relatively small number of
  agents

9
Analysis concepts
10
Concepts

• System e.g. society of agents
• Roles
• E.g. president, mailman, secretary
• In an actual instantiation, these will be
  instantiated by actual persons / agents.
• An agent can fulfill multiple roles, the same
  role can be taken on by multiple agents.
• A role is defined by four attributes
• Responsibilities liveness properties safety
  properties
• Permissions resources
• Activities
• Protocols e.g. Dutch auction, English auction
  etc.

11
The roles model

• Identifies the key roles of the system.
• Here a role can be viewed as a more or less
  abstract description of an entitys expected
  functionality.
• Two things to specify
• The permissions/rights associated with the role
• A role will have associated with it certain
  permissions, related to the type and amount of
  resources that can be exploited when carrying on
  a role.
• The responsibilities of the role
• A role is created in order to do something. That
  is described here.

12
The interaction model

• There are inevitably dependencies and
  relationships between roles in a multi-agent
  organization.
• In Gaia, such links are represented in the
  interaction model.
• This model consists of a set of protocol
  definitions, one for each type of inter-role
  interaction.
• Does not specify all the protocols down to the
  every message interchanged.

13
Protocol definition

• A protocol definition consists of the following
  attributes
• Purpose brief text description of the nature of
  the interaction.
• Initiator the role who is starting it.
• Responder the role(s) with which the initiator
  interacts.
• Inputs information supplied by the initiator
  while enacting the protocol
• Outputs information supplied by/to the protocol
  responder during the interaction.
• Processing brief textual description of any
  processing happening during the protocol.

14
Analysis

• 1. Identify the roles in the system.
• Output prototypical roles model.
• 2. For each role identify and document the
  associated protocols.
• Output interaction model.
• 3. Using the protocol model as a basis, elaborate
  the roles model.
• Output fully elaborated roles model, which
  documents the key roles occuring in the system,
  their permissions and responsibilities, together
  with the protocols and activities in which they
  participate.
• 4. Iterate stages 1-3.

15
Design

• Classical design transform the abstract models
  derived during analysis into sufficiently low
  level of abstraction that it can be implemented.
• In Gaia the goal of the design process is to
  transform the analysis models (roles,
  interaction) such that traditional design
  techniques (object oriented design, usually) may
  be applied in order to implement the agents.
• How a society of agents cooperate to realize
  higher level goals
• What is required of each agent
• How the agent realizes its services is the goal
  of lower level (program level) design.

16
Models in the Gaia design process

• Agent model
• The agent types that will make up the system
• Agent instances which will be instantiated
• Agent type tree leafs are roles, agents are
  other nodes
• No inheritance
• Services model
• The main services that are required by each agent
  to realize the agents role.
• For each service define inputs, outputs,
  preconditions, postconditions
• Acquintance model
• Documents the lines of communication between
  agents.
• Does not define the messages sent

17
The Gaia design process

• Create an agent model
• Aggregate agent roles into agent types, and
  refine to form an agent type hierarchy
• Document the instances of each agent type using
  instance annotations.
• Develop a services model, by examining
  activities, protocols, and safety and liveness
  properties of roles.
• Develop an acquaintance model from the
  interaction model and agent model.