Knowledge Based Systems and
Artificial Intelligence
Intelligent Agents
Aleksandra Pizurica
Department Telecommunications and Information Processing (TELIN)
Image Processing and Interpretation (IPI) Group
Statistical Image Modeling Lab
Overview
• Agents and environments
[R&N, Chapter 2]
• Rationality
• Task environment
• Environment types
• Agent types
This presentation is based on the book of S. Russel and P. Norvig: Artificial Intelligence: A Modern Approach, (Third Edition, 2010),
denoted here as [R&N] and on the instructor’s resource page (slides of S. Russel ) http://aima.cs.berkeley.edu/
2
Agents and environments
3
Example: Vacuum cleaner world
4
Example: Vacuum cleaner world
5
Rationality
• A rational agent is one that does the right thing.
• How do we know whether it is the right thing?
- By considering the consequences of the agent behavior (i.e., the
sequence of states through which the environment goes as a result
of agent’s behavior)
• A sequence of states (through which the environment goes) is
evaluated by a performance measure
6
Rationality, contd.
7
Specifying the task environment
8
Specifying the task environment
9
Specifying the task environment
Take another example: Internet shopping agent
10
Specifying the task environment
Take another example: Internet shopping agent
11
Environment types
Solitaire
Backgammon
12
Environment types
Solitaire
Backgammon
13
Environment types
Solitaire
Backgammon
14
Environment types
15
Environment types
16
Environment types
17
Environment types
18
Environment types
19
Agent types
• Four basic types in order of increasing generality:
 simple reflex agents
 reflex agents with state
 goal-based agents
 utility-based agents
All these can be turned into learning agents
20
Simple reflex agents
21
Reflex agents with state
22
Goal-based agents
23
Utility-based agents
24
Advantages of utility-based agents
• Not the only way of acting rational.
• However, advantages in terms of flexibility and learning
• Can act rationally in two important cases where the others fail:
 Having conflicting goals
 Having several goals, none of which can be achieved with certainty
• In reality, partial observability  maximizing the expected utility
25
Learning agents
26
Summary
• An agent is something that perceives and acts in an environment.
• The agent function specifies the action taken in response to any percept
sequence.
• The performance measure evaluates the environment sequence.
• A rational agent maximizes expected performance.
• The agent program implements the agent function (designs vary in efficiency)
• In designing an agent a first step must be to specify the task environment as
fully as possible. The task environment specification includes: Performance
measure, Environment, Actuators and Sensors (PEAS)
• Environments are categorized along several dimensions: observable?
deterministic? episodic? static? discrete? single-agent?
• Basic agent types: reflex, reflex with state, goal-based, utility-based.
27