- •1. TABLE OF CONTENTS
- •2. AN INTRODUCTION TO UNIX
- •2.1 OVERVIEW
- •2.2 UNIX
- •2.2.1 Using UNIX Workstations in general:
- •2.2.2 Directories, Files, Etc.
- •2.2.3 Advanced Concepts
- •2.3 THE NETWORK
- •2.4 GOOD MANNERS
- •3. THE INTERNET
- •3.1 NETWORKS
- •3.1.1 Computer Addresses
- •3.2 NETWORK TYPES
- •3.2.1 Permanent Wires
- •3.2.2 Phone Lines
- •3.3 NETWORK PROTOCOLS
- •3.3.1 Mail Transfer Protocols
- •3.3.1.1 - Attachments
- •3.3.1.2 - Mail Lists
- •3.3.2 FTP - File Transfer Protocol
- •3.3.3 News
- •3.3.4 HTTP - Hypertext Transfer Protocol
- •3.3.5 Chat
- •3.3.6 Novell
- •3.3.7 Security
- •3.4 DATA FORMATS
- •3.4.1 HTML - Hyper Text Markup Language
- •3.4.1.1 - Publishing Web Pages
- •3.4.2 URLs
- •3.4.3 Hints
- •3.4.4 Specialized Editors
- •3.4.6 Encryption
- •3.4.7 Compression
- •3.5 PULLING ALL THE PROTOCOLS AND FORMATS TOGETHER WITH BROWSWERS
- •3.6 OTHER STUFF
- •3.6.1 Clients and Servers
- •3.6.2 Java
- •3.6.3 Javascript
- •3.6.5 Searches
- •3.6.6 ActiveX
- •3.6.7 Graphics
- •3.6.8 Animation
- •3.6.9 Video
- •3.6.10 Sounds
- •3.6.11 Other Program Files
- •3.6.12 Fancy Stuff
- •4. TEACHING WITH THE INTERNET
- •4.1 LECTURES
- •4.1.1 Equipment
- •4.1.2 Techniques
- •4.2 ON-LINE NOTES
- •4.3 ON-LINE MARKING
- •4.3.1 Web Pages
- •4.3.2 email
- •4.4 The Time-Line For My First On-Line Course (Fall 1996)
- •5. WWW and HTML
- •5.1 Why Bother?
- •5.2 Where to Find Netscape
- •5.3 How to Get Your Own Home Page
- •5.4 How to Create a file
- •5.5 Resources
- •6. A BASIC INTRODUCTION TO ‘C’
- •6.2 BACKGROUND
- •6.3 PROGRAM PARTS
- •6.4 HOW A ‘C’ COMPILER WORKS
- •6.5 STRUCTURED ‘C’ CODE
- •6.6 ARCHITECTURE OF ‘C’ PROGRAMS (TOP-DOWN)
- •6.7 CREATING TOP DOWN PROGRAMS
- •6.8.1 Objectives:
- •6.8.2 Problem Definition:
- •6.8.3 User Interface:
- •6.8.3.1 - Screen Layout (also see figure):
- •6.8.3.2 - Input:
- •6.8.3.3 - Output:
- •6.8.3.4 - Help:
- •6.8.3.5 - Error Checking:
- •6.8.3.6 - Miscellaneous:
- •6.8.4 Flow Program:
- •6.8.5 Expand Program:
- •6.8.6 Testing and Debugging:
- •6.8.7 Documentation
- •6.8.7.1 - Users Manual:
- •6.8.7.2 - Programmers Manual:
- •6.8.8 Listing of BeamCAD Program.
- •6.9 PRACTICE PROBLEMS
- •7. GUI DESIGN
- •7.1 PRACTICE PROBLEMS
- •8. AN EXAMPLE - BEAMCAD
- •9. PROGRAMMING IN JAVA
- •9.1 OVERVIEW
- •9.2 THE LANGUAGE
- •9.3 OBJECT ORIENTED PROGRAMMING
- •9.4 REFERENCES/BIBLIOGRAPHY
- •10. DATABASES
- •11. MESSAGE PASSING ON NETWORKS
- •12. MATHEMATICAL ELEMENTS OF COMPUTER GRAPHICS
- •12.1 INTRODUCTION
- •12.2 PIXELS
- •12.2.1 The Perspective Transform
- •12.3 LINE DRAWING
- •12.3.1 Hidden Lines
- •12.4 POLYGON DRAWING
- •12.5 SHADED POLYGONS
- •12.6 COLORS
- •12.6.1 Color Maps
- •12.6.1.1 - Quantization with an Octree RGB Cube
- •12.6.1.1.1 - Algorithm and Implementation
- •12.6.1.1.2 - Color Quantization Data Structures
- •12.7 DITHERING
- •12.7.1 A Model for Light Ray Reflection
- •12.7.2 A Model for Light Ray Refraction:
- •12.7.3 A Model for Specular Reflection of Point Light
- •12.8 RAY TRACING
- •12.8.1 Basic Ray Tracing Theory
- •12.8.1.1 - A Model for Diffuse Reflection of Ambient Light
- •12.8.1.2 - A Model for Diffuse Reflection of Point Light:
- •12.8.1.3 - Collision of a Ray with a Sphere:
- •12.8.1.4 - Collision of a Ray With a Plane:
- •12.8.1.5 - Mapping a Pattern
- •12.8.2 Ray Tracer Algorithms
- •12.8.3 Bounding Volumes
- •12.8.4 Shadows
- •12.8.5 Aliasing
- •12.8.6 Advanced topics
- •12.9 RADIOSITY
- •12.10 ADVANCED GRAPHICS TECHNIQUES
- •12.10.1 Animation
- •12.11 REFERENCES
- •12.12 PRACTICE PROBLEMS
- •13. NEW TOPICS
- •13.1 VIRTUAL REALITY
- •13.2 MULTIMEDIA
- •14. VISIONS SYSTEMS
- •14.1 OVERVIEW
- •14.2 APPLICATIONS
- •14.3 LIGHTING AND SCENE
- •14.4 CAMERAS
- •14.5 FRAME GRABBER
- •14.6 IMAGE PREPROCESSING
- •14.7 FILTERING
- •14.7.1 Thresholding
- •14.8 EDGE DETECTION
- •14.9 SEGMENTATION
- •14.9.1 Segment Mass Properties
- •14.10 RECOGNITION
- •14.10.1 Form Fitting
- •14.10.2 Decision Trees
- •14.11 PRACTICE PROBLEMS
- •15. SIMULATION
- •15.1 MODEL BUILDING
- •15.2 ANALYSIS
- •15.3 DESIGN OF EXPERIMENTS
- •15.4 RUNNING THE SIMULATION
- •15.5 DECISION MAKING STRATEGY
- •15.6 PLANNING
- •15.7 NEURAL NETWORK THEORY
- •16. ARTIFICIAL INTELLIGENCE (AI)
- •16.1 OVERVIEW
- •16.2 EXPERT SYSTEMS
- •16.3 FUZZY LOGIC
- •16.4 NEURAL NETWORKS
- •16.4.1 Neural Network Calculation of Inverse Kinematics
- •16.4.1.1 - Inverse Kinematics
- •16.4.1.2 - Feed Forward Neural Networks
- •16.4.1.3 - The Neural Network Setup
- •16.4.1.4 - The Training Set
- •16.4.1.5 - Results
page 225
-10, 20 or 40 neurons in the hidden layer
-a bias input was connected to each neuron
-the layers were all fully connected
-there were runs with one, and two hidden layers
16.4.1.4 - The Training Set
•The problem is reduced using either left or right arm configurations, the solution is also constrained to elbow up or elbow down.
•Discontinuities were avoided by not training the neural network in the region above the origin. The elbow straight configuration is also a minor singularity problem.
•Training points were evenly distributed throughout the robot workspace
•Only a quarter of the robot workspace was used because of the robot symettry.
•The general protocol for training was,
-apply the desired position to the input, and train for the desired joint angles.
-When accuracy was high enough, the first correction net was trained by comparing the actual errors, and the desired values. Additional correction networks were also trained in some cases.
-the error was measured by using an RMS measure of the differences
page 226
oerror = oactual - odesired
oaverage = Σ oerror 3n
S = |
Σ (o |
error |
)2 |
||
|
|
3n |
|
|
|
|
|
|
|
|
oactual odesired oerror oaverage S
n
-Actual angle output of Neural Network
-Desired angle output of Neural Network
-Error of angle output of neural Network
-Average angle error for all joints
-R.M.S. error for all joints
-Number of training points
• A list of results are provided below,
page 227
10 hidden neurons
20 hidden neurons
40 hidden neurons
Network |
Number of |
average |
standard |
|
absolute error |
deviation |
|||
Architecture |
Connections |
|||
degrees per joint |
degrees per joint |
|||
|
|
|||
|
|
|
|
|
1 hidden layer |
73 |
3.02 |
4.22 |
|
10 wide |
||||
|
|
|
||
1 hidden layer |
143 |
2.22 |
3.39 |
|
20 wide |
||||
|
|
|
||
1 hidden layer |
286 |
|
|
|
20 wide with |
1.66 |
2.63 |
||
1 correction net |
|
|
|
|
1 hidden layer |
|
|
|
|
20 wide with |
429 |
1.27 |
2.06 |
|
2 Correction |
||||
nets |
|
|
|
|
1 hidden layer |
|
|
|
|
20 wide with |
572 |
1.19 |
1.92 |
|
3 correction |
||||
|
|
|
||
nets |
|
|
|
|
1 hidden layer |
|
|
|
|
20 wide with |
715 |
1.04 |
1.61 |
|
4 correction |
|
|
|
|
nets |
|
|
|
|
1 hidden layer |
|
|
|
|
20 wide with |
858 |
1.01 |
1.55 |
|
5 correction |
||||
|
|
|
||
nets |
|
|
|
|
1 hidden layers |
283 |
1.72 |
2.59 |
|
40 wide |
|
|
|
|
1 hidden layer |
566 |
|
|
|
40 wide with |
1.37 |
2.21 |
||
1 correction net |
|
|
|
|
|
|
|
|
|
1 hidden layer |
|
|
|
|
40 wide with |
849 |
1.20 |
1.90 |
|
2 Correction |
||||
nets |
|
|
|
|
1 hidden layer |
|
|
|
|
40 wide with |
1132 |
1.10 |
1.77 |
|
3 correction |
||||
|
|
|
||
nets |
|
|
|
|
1 hidden layer |
|
|
|
|
40 wide with |
1415 |
1.10 |
1.76 |
|
4 correction |
||||
|
|
|
||
nets |
|
|
|
|
1 hidden layer |
|
|
|
|
40 wide with |
1698 |
1.09 |
1.76 |
|
5 correction |
||||
|
|
|
||
nets |
|
|
|
|
|
|
|
|
•The results in the table were obtained for a variety of network configurations
•A visual picture of the network configurations is shown below, and on subsequent pages. These are based on a set of test points that lie in a plane of the workspace.
page 228
********** Add figure of network point locations, and test conditions
•As seen in the experimental results, there are distortions that occur near the origin, and the edges of the workspace, as would be expected with the singularities found there.
•The errors also increased near the training boundaries
********* Add in more of the results figures
16.4.1.5 - Results
•The mathematical sigularities caused by cartesian coordinates, and the +/- 180 degrees singularity could be eliminated by selecting another set of coordinates for space and the arm.
•The best results were about 1 degree RMS.
•