Chapter 2
Intelligent Agents
Introduction
• Agent:
anything that can be viewed as
perceiving its environment through actuators.
2
Introduction
 A human agent has ears, eyes, and other organs for
sensors, and hands, legs, vocal tracts for actuators.
 A robotic agent might have cameras, infrared range
finders for sensors and various motors for actuators.
 A software agent receives keystrokes, file contents,
and network packets as sensory inputs and acts on the
environment by displaying on the screen, writing files
and sending network packets.
3
Introduction
4
Agents and Environment
•
Percept: refers to agent’s perceptual inputs at
any given instant.
•
Percept sequence: the complete history of
everything the agent has ever perceived.
•
An agent’s choice of action at any given instant
can depend on the entire percept sequence
observed to date, but not on anything it hasn’t
perceived.
5
Agents and Environment
•
Internally, the agent function for an artificial
agent will be implemented by an agent
program.
•
Agent Function: an abstract mathematical description
•
Agent Program: concrete implementation, running
within some physical system.
6
Agents and Environment
•
Figure 2.2 illustrates a vacuum cleaner world
with two locations.
•
The vacuum agent perceives which square it is
and whether there is dirt in the square.
•
It can choose to move left, move right, suck up the
dirt, or do nothing.
•
Partial tabulation of this agent function is shown
in Figure 2.3.
7
Agents and Environment
8
Agents and Environment
•
The obvious question, then, is this: How should
the vacuum act?
•
In other words, what makes an agent good or bad,
intelligent or stupid?
•
How does it know not to make a mistake?
•
Should we monitor it at all times?
•
What happens if it is put in a different
environment?
9
Good Behavior: Rationality
•
Rational Agent: one that does the right thing.
 Conceptually speaking, every entry in the table for the
agent function is filled out correctly.
•
But, what does it mean to do the right thing?
•
This question is answered the old way: by considering
the consequences of the agent’s behavior.
10
Good Behavior: Rationality
•
When agent is plunked down in an environment, it
generates a sequence of actions according to the
percepts it receives.
•
If the sequence is desirable, then the agent has
performed well.
•
The notion of desirability is captured by a performance
measure that evaluates any given sequence of
environment states.
11
Rationality
•
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 us to a definition of a rational agent:
12
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.
13
Rationality
•
Is the Vacuum in Figure 2.3 a rational agent?
•
With what it knows about the environment, yes, it is a
rational agent.
•
The Vacuum would stop being rational if, when all the
dirt is cleaned up, the agent oscillates needlessly back
and forth.
14
Rationality
•
Rationality is not the same as perfection
 Cross The Street Example
•
Rationality maximizes expected performance.
 It is not perfection
•
Information gathering: doing actions in order to
modify future percepts.
•
Our definition requires a rational agent not only to
gather information, but also to learn as much as possible
from what it perceives.
•
In extreme cases in which the environment is completely
known (priori), the agent simply acts correctly.
15
16
Rationality
•
To the extent that an agent relies on the prior knowledge
of its designer rather than on its own percepts, we say
the agent lacks autonomy.
•
A rational agent should be autonomous, it should learn
what it can to compensate for partial or incorrect prior
knowledge.
17
The Nature of Environments
•
Task Environments: the “problems” to which the
rational agents are the “solution”.
•
Task Environment Specification (PEAS)
 Performance
 Environment
 Actuators
 Sensors
18
The Nature of Environments
19
The Nature of Environments
20
The Nature of Environments
•
Properties of Task Environments
 Fully Observable vs Partially Observable
 Fully Observable – sensors detect all aspects that are
relevant to the choice of action
 Partially Observable – parts of the state are missing from
the sensor data.
21
The Nature of Environments
Task Env.
Observable Agents
Chess
Fully
Poker
Partially
Image
Analysis
Fully
Butler Robot
Partially
Episodic
Static
Discrete
22
The Nature of Environments
•
Properties of Task Environments
 Single Agent vs Multi agent
 Multi Agent – many agents are in the environment that may
affect choice of actions
 Single Agent – only one agent in the environment
23
The Nature of Environments
Task Env.
Observable Agents
Chess
Fully
Multi
Poker
Partially
Multi
Image
Analysis
Fully
Single
Butler Robot
Partially
Single
Episodic
Static
Discrete
24
The Nature of Environments
•
Properties of Task Environments
 Episodic vs Sequential
 Episodic – the agent’s experience is divided into episodes. In
each episode the agent receives a percept and then performs a
single action. The next episode does not depend on the actions
taken in previous episodes.
 Sequential – current decision could affect all future decisions,
so the environment needs to be checked every time a decision is
to be made
25
The Nature of Environments
Task Env.
Observable Agents
Episodic
Chess
Fully
Multi
Sequential
Poker
Partially
Multi
Sequential
Image
Analysis
Fully
Single
Episodic
Butler Robot
Partially
Single
Sequential
Static
Discrete
26
The Nature of Environments
•
Properties of Task Environments
 Static vs Dynamic
 Static – when environment does not change while an agent is
deliberating
 Dynamic – when environment changes while deliberating
27
The Nature of Environments
Task Env.
Observable Agents
Episodic
Static
Chess
Fully
Multi
Sequential
Static
Poker
Partially
Multi
Sequential
Static
Image
Analysis
Fully
Single
Episodic
Static
Butler Robot
Partially
Single
Sequential
Dynamic
Discrete
28
The Nature of Environments
•
Properties of Task Environments
 Discrete vs Continuous
 Discrete: the environment has set attributes
 Continuous: infinite number of attributes for the environment
29
The Nature of Environments
Task Env.
Observable Agents
Episodic
Static
Discrete
Chess
Fully
Multi
Sequential
Static
Discrete
Poker
Partially
Multi
Sequential
Static
Discrete
Image
Analysis
Fully
Single
Episodic
Static
Continuous
Butler Robot
Partially
Single
Sequential
Dynamic
Continuous
30