Grand Challenge

1
Grand Challenge

• Final Presentation
• September 14, 2009
• Blake Baccigalopi
• Paul Bowling
• Chad Larsh
• Advisor Dr. Parten

2
Outline

• What is the Grand Challenge
• Obstacle Detection using 2D LADAR
• GPS Course Planning
• Road Following using Image Analysis
• Current Vehicle Status Options
• System Integration
• Budget Gantt Chart
• Where does the project go from here

3
What is the Grand Challenge
4
What is the Grand Challenge?

• The DARPA Grand Challenge is a field test of
  autonomous ground vehicles over realistic terrain
  which sets specific performance goals for
  distance and speed.

http//www.darpa.mil/grandchallenge/Rules_8oct04.p
df
5
What is our part?

• Grand Challenge Vehicle
• Power train
• Actuators
• Interface Power
• Guidance Navigation
• Sensors

6
What is our part?

• Grand Challenge Vehicle
• Power train
• Actuators
• Interface Power
• Guidance Navigation
• Sensors

7
Official Rules

• The route may include paved roads, unpaved
  roads, trails, and off-road desert areas.
  Examples of obstacles include ditches, berms,
  washboard, sandy ground, standing water, rocks
  and boulders, narrow underpasses, construction
  equipment, concrete safety rails, power line
  towers, barbed wire fences and cattle guards. The
  route can be traversed by a commercial 4X4 pickup
  truck. In addition to the existing natural
  obstacles, DARPA will place on the route static
  obstacles that may disable a vehicle if struck.
  These obstacles must be detected and
  circumnavigated for a vehicle to successfully
  complete the route. The route is wide enough for
  vehicles to bypass these obstacles.

8
Obstacle DetectionLaser Distance Ranging (LADAR)
BACCIGALOPI
9
LADAR ImagesCaracol, MexicoPhotos taken by
University of Texas
BACCIGALOPI
http//www.geo.utexas.edu/zacaton/3D_Mapping/ladar
_scanning.htm
10
LADAR ImagesZacotan, MexicoPhotos taken by
University of Texas
BACCIGALOPI
http//www.geo.utexas.edu/zacaton/3D_Mapping/ladar
_scanning.htm
11
2D LADAR
BACCIGALOPI
http//www.geo.utexas.edu/zacaton/3D_Mapping/ladar
_scanning.htm
12
2D LADARSICK LMS 291-S05
BACCIGALOPI
13
LMS 291-S052D LADAR ScannerCost 3,014
BACCIGALOPI

• Data Specifications
• Data Interface RS 232 / RS 422 (configurable)
• Transfer Rate 9.6 / 19.2 / 38.4 / 500 kBd
• Sensing Specifications
• Scanning Speed 75 Hz
• Angular View 100, 180
• Angular Resolution 0.25 / 0 .50 / 1.00
• Measurement Resolution 10 mm

• Physical Specifications
• Weight approx. 19.8 lb
• Environment Outdoor
• (Fog Correction)
• Electrical Specifications
• Supply Voltage
• 24 V DC 15

14
SICK LMS 291 Enclosure Ratings
BACCIGALOPI

• NEMA 4 Intended for indoor or outdoor use
  primarily to provide a degree of protection
  against wind blown dust and rain, splashing
  water, and hose directed water.
• IP 65 Enclosure protected against all dust
  contamination water projected by a nozzle
  against the enclosure shall have no harmful
  effect.

15
RS-232 Command and Control of the LMS 291-S05
BACCIGALOPI
16
Initial Programming
BACCIGALOPI
17
LMS 291 Defaults
BACCIGALOPI
Startup LMS Output
18
Changing Baud Rate
BACCIGALOPI
Changing Angular Range and Resolution
19
Scanning Speed
BACCIGALOPI

• 75 Hz Scanner

20
Angular Range and Resolution
BACCIGALOPI
21
Changing Measurement Mode (mm)
BACCIGALOPI
Changing Measurement Mode (cm)
22
Start Continuous Data Output
BACCIGALOPI
Stop Continuous Data Output
23
LMS Software V5.1 from SICK
BACCIGALOPI
24
LMS Software V5.1 from SICK
BACCIGALOPI
25
LMS Software V5.1 from SICK
BACCIGALOPI
26
Global Positioning System(GPS)
BACCIGALOPI
27
Garmin GPS 16A
BACCIGALOPI

• Cost 315
• Update Rate 5 Hz
• WAAS Capable
• Resistant to 6 gs
• Serial

• Baud Rates 300/600/1200/2400/4800/9600/19200
• Input Voltage 8 40 VDC
• Weight 1.1 lbs with cable
• IPX7 submerged in 1 m water for 30 min.
• Dust Resistant with Teflon coated cable

28
WAAS(Wide Area Augmentation System)
BACCIGALOPI

• Utilizes 25 ground stations across the US to
  correct GPS inaccuracies

• Developed by FAA and DOT.
• Accurate to less than 3 m 95 of the time.

29
Last Years RDDF
LARSH
30
Road Following using Image Analysis
31
Road Following/Path Determination
BOWLING

• Objectives
• Define desired paths properties
• Could be a road
• Detect properties with sensors
• Analyze sensor output to make decisions

32
Properties of Roads
BOWLING

• Height
• low in comparison to vehicle
• Possibly higher/lower than surroundings
• Smoothness
• Variance of height
• Different from grass and rocks

33
Properties of Roads
BOWLING

• Color
• Color variance could be small
• Edges may have distinct color differences
• Different from surroundings
• Texture
• Ruts, tracks, other directional textures
• Different from surroundings

34
Properties of Roads
BOWLING

• Heat
• Paved road possibly hotter than dirt or grass
  surroundings
• Dirt road may be cooler than rocky surroundings
• Needs to be further studied

35
How to Detect Properties
BOWLING

• Laser range finding
• Formation of 3D range map (height, smoothness)
• Video Camera
• Edge detection (intensity difference)
• Color processing (color edges, color variance)

36
Edge Detection
BOWLING

• MATLAB
• Derivative of intensity of grayscale image
• Maxima and minima of derivative show
  color/intensity boundaries (edges)

37
Edge Detection
BOWLING

• Video -gt Frame -gt Image -gt Edge detect

38
Color Processing
BOWLING

• MATLAB
• RGB2IND changes image format and can reduce
  number of colors
• Variance of color can be found
• Edges of colors can be found

39
Starting Point
BOWLING
40
Reduce Colors
BOWLING
41
Eliminate Image Above Horizon
BOWLING
42
Color Variance
BOWLING
43
Find 2 major colors
BOWLING
44
Median Filter
BOWLING
45
Color Edges
BOWLING
46
BWLabel
BOWLING
47
Selected Road Object
BOWLING
48
Middle Point of Road Object
BOWLING
49
Path Through Road
BOWLING
50
Do we need to follow a road?
BOWLING

• No
• GPS gives intended direction
• Road can be ignored (or given low priority) if
  heading wrong direction
• Smooth, off road path can still be found

51
Current Vehicle Status
52
Current Vehicle Status
LARSH

• Information pertaining to current
• Location
• Heading
• Tilt Angle
• Acceleration
• Speed

53
Options
LARSH

• GPS/INS Sensor
• Advantages
• All in one solution
• More accurate
• Disadvantages
• Very expensive (5,000 - 50,000)
• Design our own system
• Advantages
• Less expensive
• Disadvantages
• Time

54
GPS/INS System
LARSH

• Position - 3.9m
• Velocity - 0.5 m/s
• Attitude - 1.0 mrad
• Time - 1 µs
• Heading - 1.5 mrad
• Cost 25,000

http//www.systron.com/pro_CMIGITS.asp
55
Design our own
LARSH

• Garman 16A
• 5 Hz, 3m, Rugged, 315
• Magnetic Compass
• 3-4, 0.1, 50
• Model 900 Biaxial Clinometer
• 100 range, 0.02, 225

http//www.gps4fun.com/gar_gps16a.php http//www.h
obbyengineering.com/SectionS.html http//www.geome
chanics.com/dspproduct.cfm?prid6
56
System Integration
57
System Integration
LARSH

• Many sensors and sensor subsystems
• Information needs to be synthesized
• Course needs to be selected
• Translated to the steering throttle

58
Methods of Calculating Course
LARSH

• Strict IF statements
• Fuzzy Logic
• Neural Networks

59
Fuzzy Logic
LARSH

• Classifies strict values into fuzzy categories
• Makes decisions based on fuzzy categories
• Has the ability to assign weights

60
How does Fuzzy logic work?
LARSH
61
Neural Networks
LARSH

• Operates on a system of weights biases
• Most systems can be modeled by a series of neural
  networks
• System can be trained
• If a solution exists, MATLAB claims it will find
  it within a finite number of iterations.

62
Neural Networks
LARSH
MATLAB Neural Network Toolbox
63
Simulator
LARSH

• Need a method to test system integration
  algorithms
• Independent of sensor types
• Assumes the input information and calculates a
  desired course

64
Simulator
LARSH

• Implemented in MATLAB
• Loads bitmap image maps
• Using the test algorithm calculates a course

65
The First Implementation
LARSH

• Looks at 8 pixels around
• Calculates desired direction
• If no object in the way, it moves the vehicle
  that direction

66
Second Implementation
LARSH

• Keeps track of what direction the vehicle is
  facing
• Only looks at the 3 pixels in front
• Looks at terrain, roads, and objects
• Calculates route
• If gets stuck, backs up and tries again

67
Simulator Example
LARSH
68
Simulator Example
LARSH
69
Simulator Example
LARSH
70
LARSH
71
Budget Gantt Chart
72
LARSH
73
Gantt Chart
74
Gantt Chart
75
Gantt Chart
76
Gantt Chart
77
Where does the project go from here

• Troubleshoot LADAR System
• Translating image analysis to real time
• Acquire a current vehicle status solution
• Simulate other methods of system integration
• Incorporate system with the selected vehicle

78
Questions