Intelligent Agent

1
Chapter 6

• Intelligent Agent

2
What is Software Agent? A definition

• A computer program that carries out tasks on
  behalf of another entity. Frequently used to
  reference a program that searches the Internet
  for information meeting the specified
  requirements of an individual user.
  library.csun.edu/mwoodley/dublincoreglossary.html
  
• Computing A "smart" computer program (or
  infomachine) that can "serve" its human "master"
  in cyberspace. Software agents (sometimes also
  known as bots protect their users from the
  complexity of computer and network operations,
  and may engage in database searches and
  transactions based upon a knowledge of an
  evolving user profile. For more information see
  the UMBA Agents Web. www.nottingham.ac.uk/cyber/f
  ullglos.html
• A component of software and/or hardware which is
  capable of acting in order to accomplish tasks on
  behalf of its user. Software agents are agents in
  the form of programs (code) which operate in
  computer environments. its-proxy1.massey.ac.nz/is
  157730/year_2001/group2/glossary.htm
• agent as an autonomous, (preferably) intelligent,
  collaborative, adaptive computational entity.
  Here, intelligence is the ability to infer and
  execute needed actions, and seek and incorporate
  relevant information, given certain goals.
• http//www-2.cs.cmu.edu/softagents/intro.h
  tm

3
So, what actually agent?

• An autonomous agent is a system situated within
  and a part of an environment that senses that
  environment and acts on it, over time, in pursuit
  of its own agenda and so as to effect what it
  senses in the future.

4
Where did agents come from

• 1970s
• Comp. used primarily as engine to drive large
  applications -gt replaced human
• Comp. was good in calculation but really bad in
  communication
• Hundreds of staffs were replaced by dozens of
  staffs -gt who punched data into a format could be
  readily absorbed by the computer.

5
Where did agents come from (cont..)

• 1980s
• Focus changed, computer began to reach out
• The emphasis was on information capture at source
  -gt Information System
• Another cost saving justification
• A number of efficiencies were introduced
• Accuracy was improved as data was reaching the
  computer directly not via re-keying
• Computer capable to immediately alert when any
  problem with information by checking it as it was
  entered
• Information became timelier as there was less
  delay between capture and analysis

6
Where did agents come from (cont..)

• 1990s
• Continued with the quality of screen-gt age of GUI
• Began to be linked together using network-gt
  support file and print sharing
• Period categorized as Information Technology (IT)
• Starting with simple database-gtadvanced
  technology such as data warehouse, data mining-gt
  offer new ways to serendipitously access and
  correlate information

7
Where did agents come from (cont..)

• New Century, Millennium era
• Combining the talents of computer to calculate
  with speed and network for communicating -gt
  technology of filtering
• Filtering -gt intelligent agents will bring the
  values

8
Why do we need agents

• By doing the searching and soon doing even the
  negotiation
• Decision support
• Frontline decision support
• Repetitive office activities
• Mundane personal activity
• Search and retrieval
• Domain experts

9
Why do we need agents?

• Agents can improve the productivity of the end
  user by performing a variety task -gt most
  important to gathering information, filtering it
  and using it for decision support

10
How agents differ from conventional software
technology

• autonomy agents react themselves to observations
  of their environment without requiring explicit
  commands,
• Proactiveness/ Inteligence agents recognize and
  react to changes in the environment which present
  opportunities,
• Embeddedness/ Ability to Learn   agents' actions
  respect the real-time constraints imposed by the
  environment,
• Distributedness / Cooperation   many different
  kinds of agents can work together in the same
  system, and be added or removed without
  interrupting it.

11
Agent Classifications

• Agents can be classified according to a number of
  parameters.
• The main classes of agents are
• reactive agents
• Collaborative agents
• Interface agents
• Mobile agents
• Information-gathering agents

12
Reactive Agents
Agent Classifications Contd

• Also known as a reflex agent - a production
  system where inputs from the environment are
  compare with rules to determine which action to
  carry out
• Example automatic mail filter filter each
  e-mail and compare it against a set of rules and
  classifies it accordingly.
• Cannot work effectively when the environment
  changes example receive e-mail entirely in
  chinese.

13
Reactive Agents Contd
Agent Classifications Contd

• There are three advance type
• Goal-based Agents
• Utility-based Agents
• Utility Agent

14
Reactive Agent Goal-based Agents
Agent Classifications Contd

• More complex than reactive agents
• Works by follow a predetermined set of rule and
  try to archive a goal done by using search an
  planning
• Example goal of finding page on the internet
• agent was designed capable to carrying out
  actions (loading a web pages, examining it, and
  following links) also able to identify when it ha
  reached a goal.
• Although, to reached a goal, this agent does not
  take into account about how well it has satisfied
  the goal

15
Reactive Agent Utility-based Agents
Agent Classifications Contd

• Similar to based-goal agent, but in addition to
  achieves a set of goals, it also trying to
  maximise some utility value (happiness of the
  agents/ how successful it is being).
• Example search for page on the internet use
  knowledge about the internet to follow the most
  worthwhile paths in other word, use
  heuristic-based search technique minimize
  amount of time spends examining pages that not
  interest.

16
Reactive Agents Utility Function
Agent Classifications Contd

• Map a set of states to the set of real numbers
  this agent ale to use its utility function to
  drive a score or utility value
• Example agent might assign a high utility value
  to pages that are written in English (reliable
  source)
• The idea of utility is closely related to the
  idea of rationality

17
Collaborative Agents
Agent Classifications Contd

• Multiagent systems in which the agents
  collaborative with each other to accomplish goals
• Typically do not have the ability to learn
• Advantage of parallel nature in order to solve
  problems faster than would otherwise be possible
• More reliable than traditional systems because
  additional agents can be added to provide
  redundancy

18
Interface Agents

• Can be thought of as a personal assistant.
• Typically autonomous agents, capable of learning
  in order to carry out tasks on behalf of a human
  user.
• Collaborate with the user but no need to
  collaborate with other agents
• Example a tool to help user learn to use new
  software package.
• Can learn how to carry out the task by observing
  the user and then is able to repeat the task as
  required

19
Mobile Agent
Agent Classifications Contd

• Agent capable of moving from one place to another
• Case of mobile robots moving in physical
• Case of mobile s/ware agents, usually refers to
  the internet / other network
• Travel from one computer to another, gathering
  information performing actions as need
• Computer virus can be thought as a form of mobile
  agent
• The main advantage efficiency large amount of
  bandwidth can be safe.
• Another advantage, they can carry out their tasks
  asynchronous can be recalled and off on a
  particular task.

20
Information Agent
Agent Classifications Contd

• Also know as information-gathering agents
  usually used on the Internet and sometime called
  Internet Agent
• Help user find, filter and classify information
  from the vast array of sources available on the
  Internet
• Can be static/ mobile capable learn/ fixed
  collaborative / independently
• Know how to search the internet numbers of
  search tools cover as much content as possible
  and maximize their recall

21
Information Agent Contd
Agent Classifications Contd

• The real challenge precision heavily
  dependent on the ability of the agent to receive
  input instruction from the user
• Advantage- speed and able to examine pages
  asynchronously, delivering results to user
  e-mail
• Able to monitor the browsing habits of users to
  identify the kinds of material they are
  interested and to improve the performance of
  futures search

22
Agent Architecture

• There are a number of architecture that can be
  used to build intelligent agents
• Architecture of an agent the way in which
  various processing modules are connected together
  and connected to the environment
• Architecture that will be discuss are
• Subsumption Architecture
• BDI Architectures

23
Subsumption Architecture

• Architecture for intelligent agents invented by
  Brooks in 1985.
• Consists of a set of inputs, outputs and modules
  in layers. For example
• Each module is an AFSM (Augmented Finite State
  Machine) based on production rules of the form
  input -gt action.

24
Subsumption Architecture Contd

• The rules are situated action rules, as they
  determine what the agent will do in given
  situations.
• Such an agent is said to be situated.
• An AFSM triggers when its input exceeds a
  threshold.
• The layers in the architecture act
  asynchronously, but can affect each other.
• One layer can suppress the outputs of some
  layers, while taking into account output from
  other layers.

25
Subsumption Architecture Contd

• Rules situated/ situation action rules
  because of they map situations to action.
• A simple subsumption architectures was proposed
  by Brooks as control mechanism for a robot.
• Each layer design to handle one type of
  behavior
• Explore able to suppress instructions from the
  wander module to ensure that the robot continues
  to explore new territory
• Wander will take the instruction generated by
  the avoid obstacles and suppress the instruction
  in order to ensure while avoiding obstacles, the
  robot still wander around
• Avoid obstacles generate he instructions

26
BDI Architectures

• Belief Desire Intention Architectures.
• Beliefs statements about the environment.
• Desires goals
• Intentions plans for how to achieve the goals.
• The agent considers the options available, and
  commits to one.
• This option becomes the agents intention.
• Agents can be bold (carries out its intentions no
  matter what) or cautious (constantly reassesses
  its intentions).