Download Intelligent vs Classical Control

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts
no text concepts found
Transcript
Intelligent vs Classical Control
Bax Smith
EN9940
Today’s Topics





Distinguishing Between Intelligent and
Classical Control
Methods of Classical Control
Methods of Intelligent Control
Applications for Both Types of Control
Discussion
Distinguishing b/w Intelligent and
Classical Control
Classical Control

The Mathematicians Approach
–

Rigidly Modeled System
Software does what it is told
–
Intelligence comes from the Designer
Intelligent Control

The Lazymans Approach
–

System not Rigidly Modeled
Software does what it wants to
–
Intelligence comes from the Software
Shifting Intelligence
Classical Control
Designer
Software
Increasing Intelligence
Designer
Intelligent Control
Software
Methods for Classical Control
Open-Loop Control System
Closed-Loop Control System
System Modeling
First-Order System:
Second-Order System:
Classical Control Examples




PID Control
Optimal Control
Discrete-Event Control
Hybrid Control
PID Control

Proportional Control
–

Integral Control
–

Pure gain adjustment acting on error signal
Adjust accuracy of the system
Derivative Control
–
Adjust damping of the system
PID Control
t
de(t )
m(t )  K p e(t )  K I  e( )d  K D
dt
0
KI
GC (s)  K p 
 KDs
s
Optimal Control (LQR)
Optimal Control (LQR)
Inverted Pendulum
Inverted Pendulum Model
Methods for Intelligent Control
Intelligent Control Examples






Fuzzy Logic Control
Neural Network Control
Genetic Programming Control
Support Vector Machines
Numerical Learning
COMDPs - POMDPs
No System Modeling

Software learns system model
Fuzzy Logic Control

Multi-valued Logic
–
–

Rather warm/pretty cold vs hot/cold
Fairly dark/very light vs Black/White
Apply a more human-like way of thinking in the
programming of computers
Sets


Set A = {set of young people} = [0,20]
Is somebody on his 20th birthday young and
right on the next day not young?
Fuzzy Sets
Fuzzy Example – Inverted
Pendulum
Fuzzy Rules



If angle is zero and angular velocity is zero
then speed shall be zero
If angle is zero and angular velocity is pos. low
then speed shall be pos. low
…
Actual Values
Neural Network Control

Mimic Structure and Function of the Human
Nervous System
Biological Neurons

Dendrites
–
–

Synapses
–

Connects neurons
Modify signals
Connects Dendrites
Neuron
–
–
Emits a pulse if input
exceeds a threshold
Stores info in weight
patterns
Mathematical Representation of a
Neuron
Back-Propagation Neural Network
Training a Neural Network

Analogous to teaching a child to read
–
–
–

Present some letters and assign values to them
Don’t learn first time, must repeat training
Knowledge is stored by the connection weights
Minimize the error of the output using LMS
algorithm to modify connection weights
Genetic Programming Control

Output of Genetic Programming is another
computer program!
Genetic Programming Steps

Generate a random group of functions and terminals
(programs)
–
–


Execute each program assigning fitness values
Create a new population via:
–
–
–

Functions: +, -, *, /, etc…
Terminals: velocity, acceleration, etc…
Mutation
Crossover
Most fit
Which ever program works best is the result
Crossover Operation
Mutation Operation
Applications

In general,
–

Use Classical Control (Intelligent Control can take long to train)
If problem too complex
–
Use Intelligent Control
Discussion