CHAPTER TWO

Chapter Two

1
Intelligent Agents
compiled by Tomas U (Msc In IT)
 Agents
 An agent is anything that can be viewed as perceiving its
environment through sensors and acting upon that
environment through actuators
 Example:
 Human agent
 Sensors: eyes, ears, and other organs
 Actuators: hands, legs, mouth, and other organs
 Robotic agent
 Sensors: cameras, infrared range finders……
 Actuators: various motors

2 compiled by Tomas U (Msc In IT)

 Cont….
 Percept: agent’s perceptual input at any given instant
 Percept sequence: the complete history of everything
the agent has ever perceived
 Agent function: Describes agent’s behavior
 Maps any given percept sequence to action
 Is an abstract mathematical description

3 compiled by Tomas U (Msc In IT)

 Example
 Vacuum cleaner world

 Has two locations: square A and B

 The vacuum agent can perceive:
 Which square it is in
 Whether there is dirt in the square

4 compiled by Tomas U (Msc In IT)

 Cont…..
 It can choose:
 To move left, right
 Suck up dirt or
 Do nothing
 Agent function (partial)

5 compiled by Tomas U (Msc In IT)

 Rationality
 A rational agent is one that does the right thing
 the right action is the one that will cause the agent to be most
successful
 Therefore, we will need some way to measure success
(performance measure)
 We said an agent
 has sensors & actuators
 Act up on an environment and
 To say it is rational we need performance measure

6 compiled by Tomas U (Msc In IT)

 Cont…
 Therefore, these four thing can describe a rational agent:
 Performance measure
 Sensor
 Actuator
 The environment it is working in

7 compiled by Tomas U (Msc In IT)

 Performance measure
 embodies the criterion for success of an agent's behavior
 When an agent placed in an environment
 It generates a sequence of action
 This sequence of action causes the environment to go through a
sequence of states
 If the sequence is desirable, then the agent has performed well
 There is no one performance measure suitable for all agents
 It is better to design an objective performance measure

8 compiled by Tomas U (Msc In IT)

 Cont…
 Example: Vacuum world
 Weak performance measure
 By the amount of dirt cleaned up
 Better performance measure
 By having a clean floor

 As a general rule, it is better to design performance measures

according to what one actually wants in the environment, rather
than according to how one thinks the agent should behave.

9 compiled by Tomas U (Msc In IT)

 Rational agent
 From our discussion we can conclude that, What is rational at any given time
depends on four things:
 The performance measure that defines the criterion of success.
 The agent's prior knowledge of the environment.
 the actions that the agent can perform.
 The agent's percept sequence to date.
 This leads to a definition of a rational agent:
For each possible percept sequence, a rational agent should select an
action that is expected to maximize its performance measure, given the
evidence provided by the percept sequence and whatever built-in knowledge
the agent has.

10 compiled by Tomas U (Msc In IT)

 Cont….
 Example: the vacuum world
 The performance measure awards one point for each clean square at
each time step, over a "lifetime" of 1000 time steps.
 The "geography" of the environment is known a priori (as it was
shown above in the figure)
 The only available actions are Left, Right, Suck, and NoOp(do nothing).
 The agent correctly perceives its location and whether that location
contains dirt.
 Under these circumstances the agent is rational

11 compiled by Tomas U (Msc In IT)

 Characteristics of rational agent
 Rationality is not the same as perfection
 Definition of rationality does not require omniscience, because the
rational choice depends only on the percept sequence to date
 Should be able to learn
 With some initial knowledge from the designer
 Should be autonomous

12 compiled by Tomas U (Msc In IT)

 Task Environment
 Task environments are essentially the "problems" to which
rational agents are the "solution”
 Described by PEAS (Performance, Environment, Actuator, and
Sensor) description

13 compiled by Tomas U (Msc In IT)

 Cont..
 Example: Taxi driver agent
 Performance measure: getting to the correct destination,
minimizing fuel consumption, maximizing safety and passenger
comfort; maximizing profits….
 Environment: variety of roads, other traffic, pedestrians, stray
animals, road works, police cars, puddles, and potholes.
 Actuators: accelerator, brake….
 Sensors: Cameras, speedometer, sensors….

14 compiled by Tomas U (Msc In IT)

 Properties of Task Environment
 dimensions along which task environments can be
categorized are:
 Fully observable vs. partially observable.
 Deterministic vs. stochastic.
 Episodic vs. sequential
 Static vs, dynamic.
 Discrete vs. continuous.
 Single agent vs. multiagent.

15 compiled by Tomas U (Msc In IT)

 Fully observable vs. partially
observable.

 Fully observable: agent's sensors give it access to the

complete state of the environment
 Partially observable: parts of the state are simply missing
from the sensor data

16 compiled by Tomas U (Msc In IT)

 Deterministic vs. stochastic.

 Deterministic: the next state of the environment is

completely determined by the current state and the action
executed by the agent, otherwise it is stochastic
 Example
 Vacuum world = strategic(determjnistic
 Taxi driving: stochastic

17 compiled by Tomas U (Msc In IT)

 Episodic vs. sequential

 Episodic: the agent's experience is divided into atomic episodes.

Each episode consists of the agent perceiving and then performing
a single action. Crucially, the next episode does not depend on the
actions taken in previous episodes
 Eg= agent that spot defective part
 Sequential: the current decision could affect all future decisions
 Eg= Chess and taxi driving

18 compiled by Tomas U (Msc In IT)

 Static vs. dynamic
 Dynamic: the environment can change while an agent is
deliberating, otherwise static
 Example
 Dynamic: taxi driving
 Static: crossword puzzle

19 compiled by Tomas U (Msc In IT)

 Discrete vs. continuous.
 Discrete: has finite number of distinct states
 Eg= chess
 Continuous: unpredictable and infinite state
 Eg= taxi driving

20 compiled by Tomas U (Msc In IT)

 Single agent vs. multi agent.
 Single agent: only one agent in the environment
 Multi agent: two or more agents

21 compiled by Tomas U (Msc In IT)

 The structure of Agents
 The job of AI is to design the agent program
 agent program :implements the agent function mapping
percepts to actions
 Agent= architecture +program.
 Architecture: the physical structure on which the agent
program runs

22 compiled by Tomas U (Msc In IT)

 Agent Programs
 The four basic kinds of agent program that embody the
principles underlying almost all intelligent systems are:
 Simple reflex agents;
 Model-based reflex agents;
 Goal-based agents; and
 Utility-based agents.

23 compiled by Tomas U (Msc In IT)

 Simple reflex agent
 The simples kind
 Select actions on the basis of current percept
 Ignores the rest of percept history
 Eg= vacuum agent: decision based on the current location and on
whether that contains dirt.
 Developed based on condition-action rule
 If car-in-front-is-braking then initiate-braking
 Inefficient in complex environment

24 compiled by Tomas U (Msc In IT)

 Cont…
 will work only if the correct decision can be made on the
basis of only the current percept-that is, only if the
environment fully observable

25 compiled by Tomas U (Msc In IT)

 Model-based reflex agents
 Variant of simple reflex agent
 keep track of the part of the world it can't see now
 maintain some sort of internal state (model) that
depends on the percept history

26 compiled by Tomas U (Msc In IT)

 Goal-based agents
 Knowing about the current state of the environment is not
always enough to decide what to do.
 For example, at a road junction, the taxi can turn left, turn
right, or go straight on. The correct decision depends on where
the taxi is trying to get to
 The correct decision is based on where the taxi is trying to go
 That is, agent need to have a goal

27 compiled by Tomas U (Msc In IT)

 Cont…

28 compiled by Tomas U (Msc In IT)

 Utility-based agents
 Goals alone are not really enough to generate high-quality
behavior in most environments
 For example, there are many action sequences that will get the
taxi to its destination (thereby achieving the goal) but some are
quicker, safer, more reliable, or cheaper than others
 The agent need, besides goal, utility
 Utility: the quality of being useful

29 compiled by Tomas U (Msc In IT)

 Cont…

30 compiled by Tomas U (Msc In IT)

 Learning Agent
 Learning allows agent to operate in initially unknown
environment and to become more competent than its initial
knowledge alone might allow
 A learning mechanisms can be incorporated in four of the above
agent program types
 Learning agent can be divided into four conceptual components:
learning element, performance element, problem generator and
critic

31 compiled by Tomas U (Msc In IT)

 Cont..
 Learning element
 Responsible for making improvements
 Performance element
 Takes in percepts and decides on actions
 Critic
 Give feedback to the learning element how the agent is doing so that
the learning element can adjust performance element to do better in
the future
 Problem generator
 Suggest action for future improvement

32 compiled by Tomas U (Msc In IT)

 Cont…..

33 compiled by Tomas U (Msc In IT)