1 intelligent agents: software for the real world zagent: situated in an environment zreal world:...

8
1 Intelligent Agents: Software for the Real World Agent: situated in an environment Real world: dynamic, uncertain, non-deterministic ... “The best laid plans …”: things will go wrong Hence an intelligent agent is autonomous, proactive, flexible, robust, reactive, cooperative www.cs.rmit.edu.au/agents

Upload: clarissa-lloyd

Post on 17-Jan-2016

215 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: 1 Intelligent Agents: Software for the Real World zAgent: situated in an environment zReal world: dynamic, uncertain, non-deterministic... z“The best laid

1

Intelligent Agents: Software for the Real World

Agent: situated in an environmentReal world: dynamic, uncertain,

non-deterministic ...“The best laid plans …”: things will go wrongHence an intelligent agent is autonomous,

proactive, flexible, robust, reactive, cooperative

www.cs.rmit.edu.au/agents

Page 2: 1 Intelligent Agents: Software for the Real World zAgent: situated in an environment zReal world: dynamic, uncertain, non-deterministic... z“The best laid

2

Some Applications of Agents

Sequencing aircraft landings at Sydney airport

Diagnosing faults on the space shuttleRobocup (robotic soccer)Information managementBusiness process managementeCommerceHolonic manufacturing

Page 3: 1 Intelligent Agents: Software for the Real World zAgent: situated in an environment zReal world: dynamic, uncertain, non-deterministic... z“The best laid

3

The BDI Model

Beliefs - local knowledge baseDesires- what the agent is trying to

achieveIntentions - currently “adopted” plansPlans - predetermined sequences of

actions (or sub-goals) that can accomplish specified tasks

BDI model combines psychologically based ideas, formal logic, architecture, implementations and applications

Page 4: 1 Intelligent Agents: Software for the Real World zAgent: situated in an environment zReal world: dynamic, uncertain, non-deterministic... z“The best laid

4

But ...

BDI difficult to understand/useGap between theory and practiceIn practice more like beliefs,

events, plans (BEP)Even worse: E.G. JACK=

BEP+Java (i.e. Object oriented concepts, imperative programming language, expressions, types, concurrency …)

Also terms are overloaded: “desire” means different things

Page 5: 1 Intelligent Agents: Software for the Real World zAgent: situated in an environment zReal world: dynamic, uncertain, non-deterministic... z“The best laid

5

Our Goal:

To find a simplified model which retains the power of the BDI model but allows more people to develop intelligent agent systems.

SAC: Simplifying Agent Concepts

(Jan. 2001 - Dec. 2003)

Page 6: 1 Intelligent Agents: Software for the Real World zAgent: situated in an environment zReal world: dynamic, uncertain, non-deterministic... z“The best laid

6

Our Goal (Continued)

More details: Identify core concepts and provide clear

semantics Develop a simple methodology for

developing agent systems Develop a prototype environment and

support tools based on JACKTarget audience: professional developers or

senior undergrads

Page 7: 1 Intelligent Agents: Software for the Real World zAgent: situated in an environment zReal world: dynamic, uncertain, non-deterministic... z“The best laid

7

Agent Oriented Software Pty. Ltd. (AOS)

AOS is our industrial partner for this research.

AOS is an Australian company based in Carlton, Melbourne which has developed the JACK Intelligent Agentstm

agent development environment. The SAC project is funded by the Australian

Research Council (ARC) and by Agent Oriented Software AOS www.agent-software.com

Page 8: 1 Intelligent Agents: Software for the Real World zAgent: situated in an environment zReal world: dynamic, uncertain, non-deterministic... z“The best laid

8

People

Associate Professor Lin Padgham (RMIT)Dr. James Harland (RMIT)Dr. Michael Winikoff (RMIT)John Thangarajah (RMIT)David Poutakidis (RMIT)Dr. Andrew Lucas (AOS) Andrew Hodgson (AOS)Paul Maisano (AOS)