WHAT IS ARTIFICIAL INTELLIGENCE?

Download Report

Transcript WHAT IS ARTIFICIAL INTELLIGENCE?

Intelligent
Agents:
an Overview
Definitions
Rational behavior: to achieve a goal
minimizing the cost and maximizing the
satisfaction.
Rational agent:
Sensors to perceive the environment
Effectors to change the environment
A processing component that
chooses how to act depending on the input,
the goal, available actions, the environment.
2
Terminology
Percept sequence : sequence of
inputs to the agent.
Performance measure: a measure of
our satisfaction with the agent.
Utility function: used by the agent to
evaluate its actions.
3
Percepts and Actions
In general, there is a mapping between
the percept sequences and the possible
actions.
The mapping can be represented
 explicitly (table look up),
 as a function (e.g. square root agent),
 as a set of rules

4
Features of an Agent I
The agent constructs the algorithm
to accomplish the task
dynamically


Knowledge base
Reasoning mechanisms
5
Features of an Agent II
The agent accumulates experience learns from its activity and adapts

Learning


Through interaction with the user, and /or
other agents
It may change its knowledge base, it may
change its reasoning procedures
6
The PAGE Model
PAGE:
Percepts
Actions
Goals
Environment
7
General Scheme of Action
Analyze percepts and update
memory
 Choose best-action based on the
new state of memory
 Update memory after the action

8
Environment Program
The environment is described in terms of
states
Basic loop:
Do until termination state:




send percept to agent
get action from agent
evaluate action and compute
performance scores
update environment
9
Types of Agents
Simple reflex agents
 Agents that keep track of the
world
 Goal-based agents
 Utility-based agents

10
Reflex Agents
External
world
IF
Condition
THEN
Action
11
Agents that Keep Track of the
World
External
world
Memory
IF
Condition
THEN
Action
12
Goal-Based Agents
Goals
IF
Condition
THEN
Choose a rule
13
Utility-Based Agents
Utility
function
Goals
IF
Condition
THEN
Choose a
rule
14
Agents and Robots


Software agents
 Intelligent agents whose input (apart
from interaction with the user) is from
the virtual computer world
Robots
 Hardware devices
 Input - from the real world
 Need more capabilities - motion, vision,
reasoning about space.
15
Research in Robotics
NASA robotics: http://robotics.jpl.nasa.gov/groups/rv/homepage.html
Brown University: http://www.cs.brown.edu/research/areas/robotics.html
Stanford Robotics: http://ai.stanford.edu/
Humanoid Robotics at MIT AI lab:
http://www.ai.mit.edu/projects/humanoid-robotics-group/
Mark Tilden's bug robots
http://www.beam-online.com/Robots/Galleria_other/tilden.html
Robogenetics project http://www.nis.lanl.gov/projects/robot/
The Artificial Intelligence Laboratory
at the Free University of Brussels:
http://arti.vub.ac.be/
16