Lecture 2: 11/4/1435 Problem Solving Agents Lecturer/ Kawther Abas [email protected] 363CS –...

9
Lecture 2: 11/4/1435 Problem Solving Agents Lecturer/ Kawther Abas [email protected] 363CS – Artificial Intelligence

Transcript of Lecture 2: 11/4/1435 Problem Solving Agents Lecturer/ Kawther Abas [email protected] 363CS –...

Page 1: Lecture 2: 11/4/1435 Problem Solving Agents Lecturer/ Kawther Abas k.albasheir@sau.edu.sa 363CS – Artificial Intelligence.

Lecture 2: 11/4/1435

Problem Solving Agents

Lecturer/ Kawther Abas

[email protected]

363CS – Artificial Intelligence

Page 2: Lecture 2: 11/4/1435 Problem Solving Agents Lecturer/ Kawther Abas k.albasheir@sau.edu.sa 363CS – Artificial Intelligence.

Problem Solving Agents

Problem solving agent– One kind of agent– Find sequences of actions leading to desirable

environment states (goal)– Defining goal, makes it easier to define performance

measure, needed in definition of rational agent Problem formulation

– What aspects of the problem do we represent?– Representing actions, states, goals– These are typically represented in a programming

language, e.g., Python

Page 3: Lecture 2: 11/4/1435 Problem Solving Agents Lecturer/ Kawther Abas k.albasheir@sau.edu.sa 363CS – Artificial Intelligence.

Problem Formulation

Actions and states to considerStates possible world statesAccessibility the agent can determine via itssensors in which state it isconsequences of actions the agent knows theresults of its actionsLevels problems and actions can be specified atvarious levelsConstraints conditions that influence theproblem-solving processPerformance measures to be appliedCosts utilization of resources

Page 4: Lecture 2: 11/4/1435 Problem Solving Agents Lecturer/ Kawther Abas k.albasheir@sau.edu.sa 363CS – Artificial Intelligence.

Defined problems and solutions

Definition of a problem: collection of information that the agent will use to decide what to do– Start state– Actions

• Can specify as successor function or as operators• Actions that can be applied to a state

– Goal test• set of properties or abstract specification• tests a state to see if we have achieved goal

– Path cost function• assigns a numeric cost to a path, this is the sum of the step

costs (e.g., operator costs) along the path

Page 5: Lecture 2: 11/4/1435 Problem Solving Agents Lecturer/ Kawther Abas k.albasheir@sau.edu.sa 363CS – Artificial Intelligence.

Goal & Start States; Actions

Goal state– A state we want to achieve

• E.g., we want to win a game of chess• E.g., we want to be close to the wall in wall following

– Set of environment states that we want achieve Also have a start state State representation

– Means of defining states (e.g., in Python) of our environment Actions

– What can we do to a state?• In chess, we can move a player, constrained by the way a player can be

moved• With a robot-agent, we can move turn, translate (move forward)

– Actions cause transitions between states– Try to transform current state into goal state

Page 6: Lecture 2: 11/4/1435 Problem Solving Agents Lecturer/ Kawther Abas k.albasheir@sau.edu.sa 363CS – Artificial Intelligence.

6

How to implement a general agent?

Page 7: Lecture 2: 11/4/1435 Problem Solving Agents Lecturer/ Kawther Abas k.albasheir@sau.edu.sa 363CS – Artificial Intelligence.

Overall Process

Search for goal

Input: start state, goal state

Computation

Output:

sequence of operators, and/or goal state

Input: actions knowledge

engineering

Des

igne

r/

engi

neer

Page 8: Lecture 2: 11/4/1435 Problem Solving Agents Lecturer/ Kawther Abas k.albasheir@sau.edu.sa 363CS – Artificial Intelligence.

Types of Problems

Generally assume environments are– Fully observable, deterministic, sequential,

static, discrete, single agent – If a problem can be solved directly or

analytically– Practically, don’t apply problem solving using

search– E.g., solving 2 + 2 = 4; solving for the roots of

a quadratic; programming a spreadsheet

Page 9: Lecture 2: 11/4/1435 Problem Solving Agents Lecturer/ Kawther Abas k.albasheir@sau.edu.sa 363CS – Artificial Intelligence.

Problem Solving Costs

Path Cost– Cost of operators along path to goal

Search Cost– Cost of expanding nodes in entire search process– E.g., cost of node expansion is 1

• search cost is number of nodes expanded in search process