intelligent agents & agent-oriented systems james harland school of computer science and it...
TRANSCRIPT
Intelligent Agents & Agent-oriented systems
James Harland
School of Computer Science and IT
Intelligent Agents & Agent-oriented systems
RMIT University© Intelligent Agents 2
RMIT University© Intelligent Agents 3
Why do we need Intelligent Systems?
• New York Stock Exchange
• 33% of shares (by volume) traded by machines
• 40% of shares (by value) traded by machines
• Needs fast response time
RMIT University© Intelligent Agents 4
Why do we need Intelligent Systems?
“The Robots are coming! The Robots are coming!”
• Mars Rover
• Home systems – vacuum, lawn mower, gutter cleaner
• Military systems --- mine clearance, sensing, weapons control, autonomous vehicles, …
• Sensor networks
• Exoskeletons and artificial limbs
• Rescue robots
• Fish, birds, dogs, insects, …
• …
RMIT University© Intelligent Agents 5
Robots
RMIT University© Intelligent Agents 6
What is Intelligence?
talk “Hello! How are you today. Could I please have a coffee?”
hear Interpret “er… so.. um… as I was saying, what? No! Coffee please!”
read Interpret “Time flies like an arrow. Fruit flies like a banana.”
think “Low on fuel, it’s getting dark, no clue where I am … Hmmmm….”
act Pick things up, move them, walk, build things, …
see Distinguish moving objects from static ones, colours, textures, …
adapt This isn’t working, so let’s try something else …
paint Make judgements about what is pleasing and what is not
RMIT University© Intelligent Agents 7
Intelligent Systems in CS & IT
• Agent oriented programming
• Software that observes, thinks, acts and reacts
• Natural language processing and generation
• Software that can have a dialogue with humans
• Computational reasoning
• Software that can work out what to do
• Machine learning
• Software than can analyse, adapt and be trained
• Evolved Art
• Artistic collaboration with School of Media and Communication
• Interactive Toys
• Collaboration with XSiVE on toys which interact with children and grow with them
RMIT University© Intelligent Agents 8
What is an agent?
situated: in an environment which can be sensed and acted onpro-active: pursues its agenda, despite the environmentreactive: responds appropriately to changes in the environmentautonomous: can act on its own, without waiting to be instructedsocial: can interact with other agents, and possibly humans ….
RMIT University© Intelligent Agents 9
Beliefs, Desires, Intentions
• Agents often described in terms of mental attitudes (Beliefs, Desires, Intentions)
• Based on philosophical analysis of human reasoning by Bratman
• Developed in the early 1990s by Rao and Georgeff (in Melbourne!)
• Many variations since then
Beliefs: What it currently understands as the state of the world
Desires: States of the world that it would like to be true
Intentions: Ways of getting from current state to desired states
RMIT University© Intelligent Agents 10
‘Observe, think, act’
How does this work?
Observe
ObserveReceive and process
information about the environment
ThinkDetermine which plan should be
running
ActExecute part of chosen plan
RMIT University© Intelligent Agents 11
Example
Vacuum cleaner has to clean three rooms
It can move, or polish, and can sense whether a room is clean or dirty.
Beliefs:
clean(1), dirty(2), dirty(3), in(1)
Goals:
clean(1), clean(2), clean(3)
Plan:
move(1,2) then polish(2) then move(2,3) then polish(3)
Observe
RMIT University© Intelligent Agents 12
Example
Vacuum cleaner has to clean three rooms
It can move, or polish, and can sense whether a room is clean or dirty.
Beliefs:
clean(1), dirty(2), dirty(3), in(2)
Goals:
clean(2), clean(3)
Plan:
polish(2) then move(2,3) then polish(3)
Observe
RMIT University© Intelligent Agents 13
Example
Vacuum cleaner has to clean three rooms
It can move, or polish, and can sense whether a room is clean or dirty.
Beliefs:
clean(1), clean(2), dirty(3), in(2)
Goals:
clean(3)
Plan:
move(2,3) then polish(3)
Observe
RMIT University© Intelligent Agents 14
Example
Vacuum cleaner has to clean three rooms
It can move, or polish, and can sense whether a room is clean or dirty.
Beliefs:
clean(1), clean(2), dirty(3), in(2)
Goals:
clean(3)
Plan:
polish(3)
Observe
RMIT University© Intelligent Agents 15
Agents
• Applied in many contexts --- robots, finance, air traffic control, simulations,movies,….
• Prometheus design methodology (like UML for agents)
• Agent-oriented software engineering
• Agent programming languages
• Model-checking for agents
• Reasoning about agents, beliefs, goals,…
• Agents that learn
• Bushfire simulation
• Simulation of populations
• …
Observe
RMIT University© Intelligent Agents 16
Questions?