3-D Scanning Robot

1
3-D Scanning Robot

• Steve Alexander
• Jeff Bonham
• John Johansson
• Adam Mewha
• Faculty Advisor Dr. C. Macnab

2
Overview of Presentation

• Project Goal
• Background of 3D Scanning
• Structural and Mechanical Design
• Software Design
• Control
• Communication
• Visualization
• Results

3
Project Goal

• Create an autonomous robot which will scan an
  object and create a 3-D representation of that
  object on a computer

4
Background of 3-D Scanning

• Basic Idea
• Create a 3-D representation of an object which
  can be viewed on a computer

5
Background of 3D Scanning

• Existing Scanning Methods
• Laser
• Photogrammetry
• Direct Contact
• Sound Waves

6
Background Mechanics

• Decided to go with a three-joint arm using a
  direct-contact method
• Using Lego for the design
• Advantages
• Low Cost
• C Programming Interface
• Wide Variety of Parts Leading to Easier
  Structural Design And Redesign

7
Background Mechanics

• Disadvantages
• Limited Number of Input And Output Connections
• Limited Processing Power
• Gear Backlash And Structural Flexibility
• Result Poor Accuracy

8
Background Mechanics

• Sample gears and gear reduction

Diagrams from D. Baum, "Definitive Guide to Lego
Mindstorms",2nd ed.
9
Mechanics of the Current Design

• Successfully implemented stated goal of designing
  a robot arm with 3 Degrees of Freedom on a linear
  tracking base

10
Kinematics/Mechanics of the Current Design
Side view representation of the finished arm.
Equation to convert the above angles and lengths
into Cartesian coordinates
11
Mechanics of the Current Design

• January 2003 identified three main areas which
  required significant changes to be able to have a
  functional unit
• Complete redesign of Wrist/End-Effector
• Redesign/manufacture of axel support
• Complete redesign of drive system

12
Wrist / End-Effector

• Problems Identified
• End wheel too large
• Length not easily adjusted
• Not enough reduction on rotation sensor

13
Wrist / End-Effector

• 1 and 2 easily fixed
• Smallest available gear used as most distal point
  (finger)
• Axel used as attachment of distal point to main
  arm

14
Wrist / End-Effector
Third design change more difficult. Tried to
implement above design (8.3X reduction).
Excessive torque at rotation sensor, wrist joint
unable to move freely.
15
Wrist / End-Effector

• Settled on this final design giving a reduction
  of 5

16
Drive System
Original Base December 2003
Original Single wheel drive December 2003
17
Drive System
New centrally mounted guide for base (left) and
robot (right)
Final design for drive system
18
Axels

• In order to support weight of Chassis employed
  10 threaded rod as axels. Attached wheels via
  threaded inserts and lock nuts.

19
Final Mechanical Design
20
Control Software

• Control Software Progression
• Version 1.1
• Turned the motors on and off, varied the speed
• Version 1.2
• Rotary encoders used as an input to control the
  stop position of each motor

21
Control Software

• Control Software Progression
• Version 1.3
• A Proportional controller was implemented. This
  enabled us to have the motor speed slow as the
  target position was approached. 

22
Control Software

• Additions in Version 1.5 to Version 2.3
• Incorporated feedback loop control for the Elbow
  joint which tracks the position of the shoulder
• Single direction communication added between RCX
  and PC
• Multiplexing functionality using external
  multiplexing board

23
Control Software

• Additions in Version 1.5 to Version 2.3
• Added functions for recording data during a scan
  pass
• Extensive calibration was carried out to optimize
  all portions of a scan

24
Control Software
Calibrate
Move Base to Next Scan Region
Reset Arm To Scan Start Position
Turn on Shoulder Motor
Calculate Elbow Position
If Scan Finished
Transmit Data to PC
25
Communication Software

• Need to transmit scan data from robot to PC for
  processing
• LegOS Networking Protocol (LNP) too complex
• Use IR program by Pavel Petrovic
• Based on small subset of LNP
• Simple unidirectional communication
• Basic CRC

26
Communication Software

• First Version
• Batch file to call IR multiple times
• Very unreliable
• Manually timed
• No flexibility

27
Communication Software

• Second Version
• IRG (IR Good) Program
• Total solution
• Fully automated
• High complexity
• Many issues
• Abandoned

28
Communication Software

• Third Version
• IRR (IR Redone)
• C equivalent of first version
• Calls IR four times with a delay
• Worked perfectly
• Only works for one scan

29
Communication Software
Listen for Start of Scan
Loop
Start
Receive Travel
Delay
Receive Wrist
Delay
Receive Elbow
Delay
Receive Shoulder
Delay
30
Communication Software

• Need a program to listen for the start of a scan
• Developed Listen Program
• Continuously read serial port
• Listen for a 4-byte start-of-scan message
• Message is simply SCAN
• Return to IRR when done

31
Communication Software
Open Com Port
Start
Loop
Check For Data
New Data
No
Delay
Yes
Read New Data
No
Search For Message
Found
Yes
Quit
32
Communication Software

• Final version (IRR Listen)
• Fully automated scan
• Errors in transmission are unrecoverable
• Must discard an entire pass

33
Visualization Software

• Matlab software
• Load in all scan files
• Compute the position of the arm based on rotation
  sensors
• Polyfit data to smooth results
• Layer scan passes
• Use Matlab mesh command

34
Visualization Software

• Raw and fitted data for shoulder, elbow, and wrist

35
Results
36
Results

• Scanning works
• Accuracy is poor
• Largely due to Lego limitations
• Polyfitting smoothes result but wrecks sharp
  edges
• Can be improved with better construction
• No 4th-dimensional data
• Time and resource constraints

37
Questions?