Peas for automated taxi driver

Artificial Intelligence Questions

Intelligent Agent

www.funFruit.com

Intelligent Agent

What is Agent?

Agents and Environments

Rational Agents

PEAS Examples:

Agent Types

Table Driven Agents

Reflex agents

Goal-based agents

Utility-based agents

Learning agents

Environment Types

www.funFruit.com

Intelligent Agent

An agent is anything that can be viewed as

perceiving its environment through sensors

and acting upon that environment through

actuators

www.funFruit.com

Intelligent Agent

Agents and environments  An agent perceives its environment through

sensors.

 The complete set of inputs at a given time is

called a percept.

 The current percept, or a sequence of

percepts can influence the actions of an agent. The agent can change the

environment through actuators or

effectors.

An operation involving an effectors is

called an action.

Actions can be grouped into action

sequences.

5

www.funFruit.com

Intelligent Agent

Agents

 Human agent:

 eyes, ears, and other organs for sensors;

 hands, legs, mouth, and other body parts for actuators

 Robotic agent:

• cameras and infrared range finders for sensors;

• various motors for actuators..

• Example AIBO entertainment robot from SONY

 A software agent:

 Keystrokes, file contents, received network packages as sensors

 Displays on the screen, files, sent network packets as actuators

6

www.funFruit.com

Intelligent Agent

Agent Example

 A simple example of an agent in a physical environment is a thermostat for a

heater.

 The thermostat receives input from a sensor, which is

embedded in the environment, to detect the temperature.

 Two states:

 temperature too cold

 temperature OK are possible.

 The first action has the effect of raising the room temperature, but this is not

guaranteed. If cold air continuously comes into the room, the added heat may

not have the desired effect of raising the room temperature.

 Each state has an associated action:

 too cold turn the heating on

 temperature OK turn the heating off.

7

www.funFruit.com

Intelligent Agent

Agent function and program

 Agent function

An agent is completely specified by the agent function

[f: P* -> A]

mapping percept sequences to actions

 Agent program

The agent program runs on the physical architecture to produce f

AGENT = ARCHITECTURE + PROGRAM

.

8

www.funFruit.com

Intelligent Agent

Vacuum-cleaner world

 Percepts: location and state of the environment, e.g., [A,Dirty], [B,Clean]

 Actions: Left, Right,

Suck, NoOp

9

www.funFruit.com

Intelligent Agent

Agent Performance

 How do you define success? Need a performance measure

 How the agent does successfully

 E.g., 90% or 30% ?

 Eg. reward agent with one point for each clean square at each time step (could penalize for costs and noise)

 The success can be measured in various ways.

 It can be measured in terms of speed or efficiency of the agent.

 It can be measured by the accuracy or the quality of the solutions achieved

by the agent

 It can also be measured by power usage, money, etc

10

www.funFruit.com

Intelligent Agent

Rational agents

 An Intelligent Agent must sense, must act, must be autonomous (to

some extent). It also must be rational.

 AI is about building rational agents

A rational agent always does the right thing.

based on what it can perceive and the actions it can perform.

Rationality depends on 4 things:

1. Performance measure of success 2. Agent’s prior knowledge of environment 3. Actions agent can perform 4. Agent’s percept sequence to date

12

www.funFruit.com

Intelligent Agent

Rational agents(cont.)

Performance measure is a criterion for success of an agent's behavior.

 E.g., performance measure of a vacuum-cleaner agent could be amount of

dirt cleaned up, amount of time taken, amount of electricity consumed,

amount of noise generated, etc.

 As a general rule,

 it is better to design performance measures according to what one actually wants in

the environment.

 Rather than according to how one thinks the agent should behave (amount of dirt

cleaned vs a clean floor)

13

www.funFruit.com

Intelligent Agent

Learning

 Does a rational agent depend on only current percept?

 No, the past percept sequence should also be used

 This is called learning

 After experiencing an episode, the agent

should adjust its behaviors to perform better for the same job next time.

www.funFruit.com

Intelligent Agent

Autonomy

 A rational agent should be autonomous- it should learn what it can to compensate for partial or incorrect prior knowledge.

 If an agent just relies on the prior knowledge of its designer rather than its own percepts then the agent lacks autonomy

 E.g., a clock

 No input (percepts)

 Run only but its own algorithm (prior knowledge)

 No learning, no experience, etc.

www.funFruit.com

Intelligent Agent

Omniscience

 Rationality is not same as omniscience.

 A rational agent is not omniscient. It does not know the actual outcome of

its actions, and it may not know certain aspects of its environment.

 Rationality requires agent to learn as much as possible.

 The rational agent has too select the best action to the best of its knowledge

depending on its percept sequence, its background knowledge and its

feasible actions.

16

www.funFruit.com

Intelligent Agent

A rational agent chooses

whichever action maximizes

the expected value of the

performance measure given

the percept sequence to date

www.funFruit.com

Intelligent Agent

Exercise

18

What are the salient features of an agent ?

www.funFruit.com

Intelligent Agent

Task Environment

PEAS

Performance measure

Environment

Actuators

Sensors

20

www.funFruit.com

Intelligent Agent

PEAS for an automated taxi driver

 Example: Agent = taxi driver  Performance measure: Safe, fast, legal, comfortable trip, maximize

profits

 Environment: Roads, other traffic, pedestrians, customers

 Actuators: Steering wheel, accelerator, brake, signal, horn

 Sensors: Cameras, sonar, speedometer, GPS, odometer, engine sensors, keyboard

21

www.funFruit.com

Intelligent Agent

PEAS for Medical diagnosis System

 Example: Agent = Medical diagnosis system

 Performance measure: Healthy patient, minimize costs, lawsuit

 Environment: Patient, hospital, staff

 Actuators: Screen display (questions, tests, diagnoses, treatments, referrals)

 Sensors: Keyboard (entry of symptoms, findings, patient's answers)

22

www.funFruit.com

Intelligent Agent

PEAS for Part Picking Robot

 Example: Agent = Part-picking robot

 Performance measure: Percentage of parts in correct bins

 Environment: Conveyor belt with parts, bins

 Actuators: Jointed arm and hand

 Sensors: Camera, joint angle sensors

23

www.funFruit.com

Intelligent Agent

PEAS for Robot Soccer Player

 Robot Soccer Player

 Performance measure: ?

 Environment - ?

 Actuators - ?

 Sensors - ?

24

www.funFruit.com

Intelligent Agent

PEAS for Interactive Mathematics tutor

agent  Interactive Mathematics tutor agent

 Performance measure: ?

 Environment - ?

 Actuators - ?

 Sensors - ?

25

www.funFruit.com

Intelligent Agent

PEAS for Internet book shopping agent

 Internet book shopping agent

 Performance measure: ?

 Environment - ?

 Actuators - ?

 Sensors - ?

26

www.funFruit.com

Intelligent Agent

Exercise

27

Write a PEAS for ATM System

www.funFruit.com

Intelligent Agent SE 420, Lecture 2

Interacting Agents

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: Freeway

www.funFruit.com

Intelligent Agent SE 420, Lecture 2

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: Freeway

www.funFruit.com

Intelligent Agent SE 420, Lecture 2

Conflict Resolution by Action Selection Agents

 Override: CAA overrides LKA

 Arbitrate: if Obstacle is Close then CAA else LKA

 Compromise: Choose action that satisfies both agents

 Any combination of the above

 Challenges: Doing the right thing

www.funFruit.com

Intelligent Agent

Agent programs

 Input for Agent Program

 Only the current percept

 Input for Agent Function

 The entire percept sequence

 The agent must remember all of them

 Implement the agent program as

 A look up table (agent function)

www.funFruit.com

Intelligent Agent

Agent types

Simple reflex agents

Model-based reflex agents

Goal-based agents

Utility-based agents

33

www.funFruit.com

Intelligent Agent

Simple reflex agents

 The agent works by finding a rule whose condition matches the current situation, as defined by perception, and then doing the action associated with the rule.

 The agent has no memory.

 Simple Reflex Agent should has condition-action pairs defining all possible condition-action rules necessary to interact in an environment

www.funFruit.com

Intelligent Agent

Problems

 Table is too big to generate and to store (e.g. taxi)

 Takes long time to build the table

 Not adaptive to changes in the environment; requires entire table to be updated if changes occur.

www.funFruit.com

Intelligent Agent

The agent should keep track of the part of the world it can't see

now. Thus agent should maintain some sort of internal state that

depends on the percept history

 Updating the internal state information requires two kinds of

knowledge to be encoded in the agent program

 Information about how the world evolves independently of

the agent

Information about how the agent's own actions affects the

world

www.funFruit.com

Intelligent Agent

information comes from sensors – percepts

Integrate percept in the state.

State evaluate the conditions- action rule in the state.

based on this, the agent choses the Action

Execute action

Update state with action

Thus a state based agent works as follows:

www.funFruit.com

Intelligent Agent

The current state of the world is always not enough to decide what to do.

It needs to add goals to decide which situations are good.

Example : car at junction ….. Need knowledge of destination

Wo r l d m o d e l ( a s m o d e l b a s e d a g e n t ) + G o a l s

A goal is a description of a desirable situation

The goal based agent act by reasoning

about which actions achieve the goal

Example

www.funFruit.com

Intelligent Agent

Goal-based agents

Goal based agents work as follows:

 information comes from

sensors – percepts

 changes the agents current

state of the world

 based on state of the world

and knowledge (memory)

and goals/intentions, it

chooses actions and does them

through the effectors

www.funFruit.com

Intelligent Agent

 What if there are many paths to the goal?

 Utility measures which states are preferable to other state

 Maps state to real number (utility or “happiness”)

 Examples

www.funFruit.com

Intelligent Agent

Learning Agent

Learning allows an agent to operate in initially unknown environments. The learning element modifies the performance element. Learning element is responsible for making improvements Performance element is responsible for selecting external actions (it is what we had defined as the entire agent before) Learning element uses feedback from the critic on how the agent is doing and determines how the performance element should be modified to do better in the future Problem generator is responsible for suggesting actions that will lead to a new and informative experiences

www.funFruit.com

Intelligent Agent

Environment types

Fully observable vs. partially observable

Deterministic vs. stochastic

Episodic vs. sequential

Static vs. Dynamic

Discrete vs. Continuous

Single agent vs. multi-agent

43

www.funFruit.com

Intelligent Agent

Environment types

Fully observable vs. partially observable

A task environment is fully observable if the sensors detect all

aspects that are relevant to the choice of action & agent can obtain

complete, timely and accurate information about the state of the

environment.

An environment might be partially observable because of noisy

and inaccurate sensors or because parts of the state are simply

missing from the sensor data.

 A local dirt sensor of the cleaner cannot tell  Whether other squares are clean or not

44

www.funFruit.com

Intelligent Agent

Environment types

Fully observable vs. partially observable

45

www.funFruit.com

Intelligent Agent

Environment types

 The environment is deterministic if the next state of the environment

is completely determined by the current state and the action executed

by the agent.

 If the environment has an element of uncertainty then the environment

is stochastic

 ATM system is deterministic or stochastic ?

Deterministic vs. stochastic

46

www.funFruit.com

Intelligent Agent

Environment types

Deterministic vs. stochastic

47

www.funFruit.com

Intelligent Agent

Environment types

 In an episodic environment, the actions of an agent depend on a number of

discrete episodes with no link between the performance of the agent in

different scenarios.

Episodic vs. sequential

48

 This environment is simpler to design

since only the current environment

needs to be considered.

 Examples: Part picking robot

 In sequential environments, the current

decision could affect all future decisions

 Examples: chess and taxi driver  Classify environment of Class

room teaching

www.funFruit.com

Intelligent Agent

Environment types

Episodic vs. sequential

49

www.funFruit.com

Intelligent Agent

Static Environment: does not change from one state to the next while the agent is considering its course of action. The only changes to the environment are those caused by the agent itself. A static environment does not change while the agent is thinking. The passage of time as an agent deliberates is irrelevant. The agent doesn’t need to observe the world during deliberation Examples crossword puzzles is static

A Dynamic Environment changes over time independent of the actions of the agent and thus if an agent does not respond in a timely manner, this counts as a choice to do nothing. Examples: taxi driving is dynamic  The environment is semi-dynamic if the environment itself does not change with the passage of time but the agent's performance score does Examples: chess when played with a clock is semi-dynamic.

Static vs. Dynamic

www.funFruit.com

Intelligent Agent

Static vs. Dynamic

NO

www.funFruit.com

Intelligent Agent

Environment types

 A limited number of distinct, clearly defined states, percepts and

actions. OR A discrete environment is one where you have finitely

many action choices, and finitely many things you can sense.

Examples:

Chess has finite number of discrete states, and has discrete set

of percepts and actions.

Taxi driving has continuous states, and

actions.

Discrete vs. Continuous

52

www.funFruit.com

Intelligent Agent

Environment types

Discrete vs. Continuous

53

www.funFruit.com

Intelligent Agent

Environment types

 An agent operating by itself in an environment is single agent

Examples: Crossword is a single agent while chess is two-agents

 Question: Does an agent A have to treat an object B as an agent or can it be treated as a stochastically behaving object .

 Whether B's behavior is best described by as maximizing a performance measure whose value depends on agent's A behavior.

Examples: chess is a competitive multi-agent environment while taxi driving is a partially cooperative multi-agent environment

Single agent vs. multi-agent

54

www.funFruit.com

Intelligent Agent

Environment types

Single agent vs. multi-agent

55

www.funFruit.com

Intelligent Agent

Summary

 In conclusion AI is a truly fascinating field. It deals with exciting but hard

problems. A goal of AI is to build intelligent agents that act so as to

optimize performance.

 An agent perceives and acts in an environment, has an architecture, and is

implemented by an agent program.

 An ideal agent(rational) always chooses the action which maximizes its

expected performance, given its percept sequence so far.

 An autonomous agent uses its own experience along with built-in

knowledge of the environment by the designer.

www.funFruit.com

Intelligent Agent

SUMMARY

 Environment Types: Environments are observable , deterministic,

episodic, dynamic, and continuous.

 PEAS(Performance measure, Environment, Actuators, Sensors)

 Agent Types

 Table Driven Agents use a percept-action table in memory to find

the next action.

 Reflex agents respond immediately to percepts.

 Goal-based agents act in order to achieve their goals.

 Utility-based agents maximize their own utility function.

 Learning agents improve their performance over time

www.funFruit.com