Artificial Intelligence

1
Artificial Intelligence

• Lecture 2
• Agents

2
Overview

• Agent and Environment
• Rationality
• World Description (PEAS)
• Task Environment Types

3
Agent Environment
4
Rational Agent

• Agent
• An entity that perceives and acts. e.g. human,
  robots, thermostat, smoke detector, etc.
• The agent function maps percept sequence to
  actions
• f P - A
• The agent function is implemented by an agent
  program, running on the agent architecture
• Rational Agent
• For any given set of environments and actions, we
  seek the agent (or class of agents) with the best
  performance

5
The Vacuum-Cleaner World

• Environment square A and B
• Percepts location and status, e.g., A,Dirty
• Actions Left, Right, Suck, NoOp

6
The Vacuum-Cleaner World
7
Vacuum-Cleaner Agent

• Mapping percept to action

function Vacuum-Cleaner-Agent(L,S) returns an
action if Sdirty, then return suck else if LA,
then return right else if LB, then return left
8
Rationality
9
Rational Agent

• An agent is rational if it always does the right
  thing
• Most successful agent
• Right and wrong is decided by the designer of the
  agent
• Needs performance measure
• The criteria that determine how successful an
  agent is
• Imposed by authority and measured in the long
  run
• Performance measure should be objective
• e.g. the amount of dirt cleaned within a certain
  time
• e.g. how clean the floor is at each time step
• Performance measure should be designed
  according to what is wanted in the environment
  instead of how the agents should behave.

10
Rational Agent

• Rationality depends on four things
• Performance measure
• Prior knowledge of the environment
• Actions
• Percept sequence to date
• Definition
• 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.

11
Performance Measure

• 100 points for each piece of dirt vacuumed up
• Minus 1 point for each action taken
• Minus 1000 points for dumping the dirt in your
  neighbors backyard

A rational agent maximizes the points given the
percept sequence Rational ? Omniscient Rational ?
Clairvoyant Rational ? Perfection Rational
Exploration, learning, autonomy
12
Environment World Description
13
Task Environment

• In order to design a rational agent, we must
  specify its task environment
• PEAS description of the environment
• To design a rational taxi agent
• Performance measure
• safety, destination, profits, legality, comfort,
  
• Environment
• US streets/freeways, traffic, pedestrians,
  weather,
• Actuators
• steering, accelerator, brake, horn,
  speaker/display,
• Sensors
• video, accelerometers, gauges, engine sensors,
  keyboard, GPS,

14
Task Environment Types
15
Task Environment types
16
Task Environment types
Fully vs. partially observable an environment is
fully observable when the sensors can detect all
aspects that are relevant to the choice of
action.
17
Task Environment types
Fully vs. partially observable an environment is
full observable when the sensors can detect all
aspects that are relevant to the choice of
action.
18
Task Environment types
Deterministic vs. stochastic if the next
environment state is completely determined by the
current state and the executed action then the
environment is deterministic.
19
Task Environment types
Deterministic vs. stochastic if the next
environment state is completely determined by the
current state and the executed action then the
environment is deterministic.
20
Task Environment types
Episodic vs. sequential In an episodic
environment the agents experience can be divided
into atomic steps where the agents perceives and
then performs A single action. The choice of
action depends only on the episode itself
21
Task Environment types
Episodic vs. sequential In an episodic
environment the agents experience can be divided
into atomic steps where the agents perceives and
then performs A single action. The choice of
action depends only on the episode itself
22
Task Environment types
Static vs. dynamic If the environment can change
while the agent is choosing an action, the
environment is dynamic. Semi-dynamic if the
agents performance changes even when the
environment remains the same.
23
Task Environment types
Static vs. dynamic If the environment can change
while the agent is choosing an action, the
environment is dynamic. Semi-dynamic if the
agents performance changes even when the
environment remains the same.
24
Task Environment types
Discrete vs. continuous This distinction can be
applied to the state of the environment, to the
way time is handled and to the percepts/actions
of the agent.
25
Task Environment types
Discrete vs. continuous This distinction can be
applied to the state of the environment, the way
time is handled and to the percepts/actions of
the agent.
26
Task Environment types
Single vs. multi-agent Does the environment
contain other agents who are also maximizing some
performance measure that depends on the current
agents actions?
27
Task Environment types
Single vs. multi-agent Does the environment
contain other agents who are also maximizing some
performance measure that depends on the current
agents actions?
28
Task Environment types

• The simplest environment is
• Fully observable, deterministic, episodic,
  static, discrete and single-agent.
• Most real situations are
• Partially observable, stochastic, sequential,
  dynamic, continuous and multi-agent.

29
Environment Classes

• Environment class
• The agent must work in different environments
• A chess program should play against a wide
  collection of humans and other programs
• Designing for a particular opponent can exploit
  specific weaknesses, but is not good for general
  play
• The performance of an agent is averaged over the
  environment class
• The agent is not allowed to consult the
  environment program.

30
Summary

• Agent and Environment
• Rationality
• World Description
• Environment Types