artificial intelligence: chapter 2 week 2 and 3

7/25/2019 Artificial Intelligence: Chapter 2 Week 2 and 3

1/43

ARTIFICIAL INTELLIGENCE

AGENTS

Dr. Zeeshan Bhatti

BSSW-PIV

Chapter 2

Institute of Information and Communication TechnologyUniversity of Sindh, Jamshoro BY: DR. ZEESHAN BHATTI 1


2/43

By: Dr. Zeeshan Bhatti

Last Time: Acting Humanly: The Full Turing Test

Alan Turing's 1950 article Computing Machinery and Intelligencediscussed conditions for considering a machine to be intelligent

Can machines think? Can machines behave intelligently?

The Turing test (The Imitation Game): Operational definition ofintelligence.

Computer needs to possess: Natural language processing, Knowledgerepresentation, Automated reasoning, and Machine learning

Problem: 1) Turing test is not reproducible, constructive, and amenable tomathematic analysis. 2) What about physical interaction with interrogatorand environment?

Total Turing Test: Requires physical interaction and needs perception andactuation.

2


3/43


Last time: The Turing Test

http://www.ai.mit.edu/projects/infolab/http://aimovie.warnerbros.com 3
http://www.ai.mit.edu/projects/infolab/http://www.ai.mit.edu/projects/infolab/


4/43



http://www.ai.mit.edu/projects/infolab/http://aimovie.warnerbros.com 4


5/43



http://www.ai.mit.edu/projects/infolab/http://aimovie.warnerbros.com5


6/43





7/43





8/43


This time: Outline

Intelligent Agents (IA) Environment types

IA Behavior

IA Structure

IA Types

8


9/43


What is an (Intelligent) Agent?

An over-used, over-loaded, and misused term.

Anything that can be viewedasperceiving itsenvironment through sensors and acting upon that

environment through its effectors or actuators tomaximize progress towards its goals.

9


10/43


A human agent has eyes, ears, and other organs for sensors andhands, legs, vocal tract, and so on for actuators.

A robotic agent might have cameras and infrared range finders forsensors and various motors for actuators.

A software agent receives keystrokes, file contents, and networkpackets as sensory inputs and acts on the environment bydisplaying on the screen, writing files, and sending networkpackets.

By: Dr. Zeeshan Bhatti 10


11/43



12/43

Agents and environments

We use the term percept to refer to the agents perceptual inputsat any given instant.

An agents percept sequence is the complete history ofeverything the agent has ever perceived.

Mathematically speaking, we say that an agents behaviour isdescribed by the agent function that maps any given perceptsequence to an action.

Internally, the agent function for an artificial agent will be

implemented by an agent program. It is important to keep these two ideas distinct. The agent function

is an abstract mathematical description; the agent program is aconcrete implementation, running within some physical system.



13/43

Agents and environments

The agent function maps from percept histories to actions:

[f: P* A]

The agent program runs on the physical architecture to produce f

agent = architecture + program


14/43

Example: Vacuum-cleaner world

This particular world has just two locations: squares A and B.

The vacuum agent perceives which square it is in and whetherthere is dirt in the square. It can choose to move left, move right,suck up the dirt, or do nothing.

Percepts: location and contents, e.g., [A,Dirty]

Actions: Left, Right, Suck, NoOp

One very simple agent function is the following: if the currentsquare is dirty, then suck; otherwise, move to the other square.


15/43

Looking at Figure 2.3, we see that various vacuum-world agents can bedefined simply by filling in the right-hand column in various ways.

The obvious question, then, is this: What is the right way to fill out thetable? In other words, what makes an agent good or bad, intelligent orstupid? We answer these questions in the next section.



16/43



PAGE (Percepts, Actions, Goals, Environment)

Task-specific & specialized: well-defined goals

and environment The notion of an agent is meant to be a tool for

analyzing systems,

It is not a different hardware or new programming

languages

16


17/43


Example: Human mind as network of thousands ormillions of agents working in parallel. To produce realartificial intelligence, this school holds, we should buildcomputer systems that also contain many agents andsystems for arbitrating among the agents' competingresults.

Distributed decision-makingand control

Challenges:

Action selection: What next actionto choose

Conflict resolution

Intelligent Agents and Artificial Intelligence

sen

sors

effectors

Agency

17


18/43


Agent Types

We can split agent research into two main strands:

Distributed Artificial Intelligence (DAI)Multi-Agent Systems (MAS) (1980 1990)

Much broader notion of "agent" (1990s present)

interface, reactive, mobile, information

18


19/43


Rational Agents

EnvironmentAgent

percepts

actions

?

Sensors

Effectors

How to design this?

19


20/43


Remember: the Beobot example

20


21/43


A Windshield Wiper Agent

How do we design a agent that can wipe the windshieldswhen needed?

Goals?

Percepts? Sensors?

Effectors?

Actions?

Environment?

21


22/43


A Windshield Wiper Agent (Contd)

Goals: Keep windshields clean & maintain visibility

Percepts: Raining, Dirty Sensors: Camera (moist sensor)

Effectors: Wipers (left, right, back)

Actions: Off, Slow, Medium, Fast

Environment: Inner city, freeways, highways, weather

22


23/43


Towards Autonomous Vehicles

http://iLab.usc.edu

http://beobots.org23
http://ilab.usc.edu/http://beobots.org/http://beobots.org/http://ilab.usc.edu/


24/43


Interacting Agents: Exercise

Collision Avoidance Agent (CAA) Goals: Avoid running into obstacles

Percepts ?

Sensors?

Effectors ?

Actions ?

Environment: Freeway

Lane Keeping Agent (LKA)

Goals: Stay in current lane

Percepts ? Sensors?

Effectors ?

Actions ?

Environment: Freeway24


25/43


Interacting Agents

Collision Avoidance Agent (CAA) Goals: Avoid running into obstacles

Percepts: Obstacle distance, velocity, trajectory

Sensors: Vision, proximity sensing

Effectors: Steering Wheel, Accelerator, Brakes, Horn, Headlights

Actions: Steer, speed up, brake, blow horn, signal (headlights)

Environment: Freeway

Lane Keeping Agent (LKA)

Goals: Stay in current lane

Percepts: Lane center, lane boundaries Sensors: Vision

Effectors: Steering Wheel, Accelerator, Brakes

Actions: Steer, speed up, brake

Environment: Freeway25


26/43


Conflict Resolution by Action Selection Agents

Override: CAA overrides LKA

Arbitrate: if Obstacle is Close then CAAelse LKA

Compromise: Choose action that satisfies bothagents

Any combination of the above

Challenges: Doing the right thing

26


27/43

GOOD BEHAVIOR:THE CONCEPT OF RATIONALITY



28/43

Rational agents

Rational Agent: For each possible perceptsequence, a rational agent should select anaction that is expected to maximize itsperformance measure, given the evidence

provided by the percept sequence and whateverbuilt-in knowledge the agent has.


29/43

What is Rational Agent?

A rational agent is one that does the right thingconceptually speaking, every entry in the table for theagent function is filled out correctly.

Obviously, doing the right thing is better than doing thewrong thing, but what does it mean to do the rightthing?

by considering the consequences of the agentsbehaviour.



30/43

Rational agents

Rationality is distinct from omniscience (all-knowing withinfinite knowledge)

Agents can perform actions in order to modify futurepercepts so as to obtain useful information (information

gathering, exploration)

An agent is autonomous if its behavior is determined byits own experience (with ability to learn and adapt)


31/43

Rational agents: Performance Measure

An agent should strive to "do the right thing", based onwhat it can perceive and the actions it can perform. Theright action is the one that will cause the agent to bemost successful.

Performance measure: An objective criterion forsuccess of an agent's behavior in any given sequence ofenvironment states.

E.g., performance measure of a vacuum-cleaner agentcould be amount of dirt cleaned up, amount of timetaken, amount of electricity consumed, amount of noisegenerated, etc.


32/43

As a general rule, it is better to design performancemeasures according to what one actually wants in theenvironment, rather than according to how one thinksthe agent should behave



33/43

Rationality?

What is rational at any given time depends on four things:

The performance measure that defines the criterion of success.

The agents prior knowledge of the environment.

The actions that the agent can perform.

The agents percept sequence to date.

This leads to a definition of a rational agent:

For each possible percept sequence, a rational agent should select anaction that is expected to maximize its performance measure, given theevidence provided by the percept sequence and whatever built-inknowledge the agent has.



34/43

Consider the simple vacuum-cleaner agent that cleans a square if it is

dirty and moves to the other square if not; this is the agent functiontabulated in Figure 2.3. Is this a rational agent?

That depends! First, we need to say what the performance measure is,what is known about the environment, and what sensors and actuatorsthe agent has. Let us assume the following:

The performance measure awards one point for each clean squareat each time step, over a lifetime of 1000 time steps.

Thegeography of the environment is known a priori (Figure 2.2)but the dirt distribution and the initial location of the agent are not.

Clean squares stay clean and sucking cleans the current square. The

Left and Right actions move the agent left and right except when thiswould take the agent outside the environment, in which case the agentremains where it is.

The only available actions are Left , Right, and Suck.

The agent correctly perceives its location and whether that location

contains dirt By: Dr. Zeeshan Bhatti 34


35/43

We claim that under these circumstances the agent isindeed rational; its expected performance is at least as

high as any other agents.



36/43

One can see easily that the same agent would be irrational underdifferent circumstances.

For example, once all the dirt is cleaned up, the agent will oscillateneedlessly back and forth;

if the performance measure includes a penalty of one point for each

movement left or right, the agent will fare poorly. A better agent for this case would do nothing once it is sure that all

the squares are clean.

If clean squares can become dirty again, the agent shouldoccasionally check and re-clean them if needed.

If the geography of the environment is unknown, the agent willneed to explore it rather than stick to squares A and B.



37/43

Exercise: Home Work.

Task: Let us examine the rationality of various vacuum-cleaner agentfunctions.

a. Show that the simple vacuum-cleaner agent function described inFigure 2.3 is indeed rational under the assumptions listed on page 38.

b. Describe a rational agent function for the case in which each

movement costs one point. Does the corresponding agent programrequire internal state?

c. Discuss possible agent designs for the cases in which clean squarescan become dirty and the geography of the environment is unknown.Does it make sense for the agent to learn from its experience in thesecases? If so, what should it learn? If not, why not?



38/43


The Right Thing = The Rational Action

Rational Action: The action that maximizes theexpected value of the performance measure given thepercept sequence to date

Rational = Best ?

Rational = Optimal ?

Rational = Omniscience ?

Rational = Clairvoyant ?

Rational = Successful ?

38

(Clairvoyant = Intuitive, Psychic, Telepathic)


39/43


The Right Thing = The Rational Action

Rational Action: The action that maximizes theexpected value of the performance measure given thepercept sequence to date

Rational = Best Yes, to the best of its knowledge

Rational = Optimal Yes, to the best of its abilities (incl.

Rational Omniscience its constraints)

Rational Clairvoyant

Rational Successful

39


40/43


Behavior and performance of IAs

Perception (sequence) toAction Mapping:f : P* A

Ideal mapping: specifies which actions an agent ought to take atany point in time

Description: Look-Up-Table, Closed Form, etc.

Performance measure: a subjectivemeasure tocharacterize how successful an agent is (e.g., speed,power usage, accuracy, money, etc.)

(degree of)Autonomy: to what extent is the agent ableto make decisions and take actions on its own?

40


41/43


Look up table

agent

obstacle

sensor

Distance Action

10 No action

5 Turn left 30degrees

2 Stop

41


42/43


Closed form

Output (degree of rotation) = F(distance)

E.g., F(d) = 10/d (distance cannot be less than 1/10)

42


43/43

Thankyou

Q & AReferred BookArtificial Intelligence: A Modern Approach., 3rd Edition, byStuart Russell and Peter Norvig, Prentice-Hall, 2003

For Course Slides and Handouts

web page:https://sites.google.com/site/drzeeshanacademy/Blog:

http://zeeshanacademy.blogspot.com/

Facebook:https://www.facebook.com/drzeeshanacademy
https://sites.google.com/site/drzeeshanacademy/http://zeeshanacademy.blogspot.com/http://zeeshanacademy.blogspot.com/https://sites.google.com/site/drzeeshanacademy/

artificial intelligence: chapter 2 week 2 and 3

Documents