Lecture-04 Intelligent Agents

You might also like

Download as pptx, pdf, or txt
Download as pptx, pdf, or txt
You are on page 1of 11

Intelligent

Agents
Lecture-4

Hema Kashyap

7 August 2015
Agent
 An Agent perceives its
environment through sensors
 And acts upon the environment
through actuators or effectors
 The complete set of input at a
given time is called percept
while can influence the actions
of an agent
 An agent can be looked as a
system which has goals and in
order to implement them it maps
percept sequences into actions
 Agents Performance measures:
the success can be measured in
various ways i.e. in terms of
speed, efficiency, accuracy,
quality, power usage, money etc.
7 August 2015
Ideal Rational Agent
 For each possible percept sequence ,an ideal rational agent
should do whatever action is expected to maximize its
performance measure, on the basis of the evidence provided
by the percept sequence and whatever built in knowledge the
agent has
 An Agent behavior can be based on both its experience and
built in knowledge
 A system is autonomous to the extent that its behavior is
determined by its own experience.
 Agent Program: A function that implements the agents
mapping from percept to actions. This will run on some
computing device which we call the architecture. Eg.
Software Agents (Softbots) to fly a flight simulator etc.
 Agent =Architecture+ Program
7 August 2015
Properties of Task Environments
1. Fully Observable /Partially Observable
2. Deterministic /Stochastic
3. Episodic/Sequential
4. Static/dynamic
5. Discrete/Continuous
6. Single Agent /Multi Agent
7. Cooperative
8. Strategic

7 August 2015
Structure of Agent/Agent Architecture
1) Table Based Agents
2) Simple Reflex Agent/Percept Based Agents
3) Model Based Reflex Agents
4) Goal Based Agents
5) Utility Based Agents
6) Learning Agents

7 August 2015
Table Based Agent
 In Table based agent the action is looked up from a table
based on information about the agent’s percept
 Its simply a mapping from percepts to action which is
defined by a program and implemented by a rule based
system , by neural network or by procedure
 Drawbacks:
 Table may be very large
 learning table may take long time
 Such system have little autonomy, as all actions are
predetermined

7 August 2015
Simple Reflex Agent/Percept Based Agents
 It acts according to a rule whose
condition matches the current state ,
as defined by the percept
 Such agents are designed based on
the condition action rule
 First built a general purpose
interprets for condition action rule
and then to create rules set for
specific task environments
 Eg. If (the car in front is braking)
then (initiate braking)
 They are also termed as Reactive or
Stimulus –Response agents
 Features:
 Efficient
 No internal representation for
reasoning ,inference
 No strategic planning
7 August 2015
 Percept based agents are not good for
multiple opposing goals
Model Based Reflex Agents

 It keeps track of the


current state of the
world using an
internal model. It
then chooses an
action in the same
way as the reflex
agent
 It keeps track of the
part of the world it
can’t see now
7 August 2015
Goal Based Agent
 Such agents keeps track
of the world state as
well as a set of goals it
is trying to achieve and
chooses an action that
will eventually leads to
achievement of its goals
 Search and planning is
needed to achieve agents
goal
 Decision making
involves future
consideration as goal
based

7 August 2015
Utility Based Agents
 If one world state is preferred t to another then it has higher utility for the agents
 A utility function maps a state (a sequence of states ) onto a real number, which
describes associated degree of happiness.
 A complete specification of the utility function allows rational decision in two kinds of
cases where goals are in adequate:
 It uses a model of the world along with the utility function
 That measures its preferences among state of the world.
 Conflicting Goals: i.e. only some of which can be achieved eg. Speed and safety
 Several goals: none o of which can be achieve with certainty. In such cases likelihood of
success depends upon the importance of goals
 Then it chooses the actions that leads to the best expected utility where expected utility
is computed by averaging overall possible outcome states, weighted by the probability
of the outcome.

7 August 2015
Learning Agents
 Learning allows the agent to operate in initially unknown environments
and to become more competent than its initial knowledge alone might
allow
 A learning agent can be divided into four conceptual components:
 Learning element: responsible for making improvements
 Performance Element: responsible for selecting external actions
 Critic: the learning element uses feedback from the on how the
performance element should be modified, to do better in future
 Problem Generator: responsible for suggesting actions that will lead to
new and informative experience
 Eg. Taxi driving agents learns from feedback in form of reward and
penalty.

7 August 2015

You might also like