2. University of North Dakota
• Grand Forks, ND
• Coldest Place in the
USA
• 14,906 Students
Team RAPTOR
• 6th Year in
Competition
• 10 Team Members
• All undergraduates–
ME & EE
6/9/2015 UND Lunabotics 2
Sponsors:
3. • Final Design
• Operating Overview
• Design Process
• Highlighted Features
• Control Software
• Testing Overview
• Conclusion
6/9/2015 3UND Lunabotics
5. • 100kg On-Board
Regolith Storage
• Magnesium Alloy
Construction
• Mining Depth of
50 cm
• Compact Folding
Design
• Camera Localization
System
6/9/2015 UND Lunabotics 6
12. • Main Construction
Material: Aluminum
• Shallow Mining
System
• Limited Storing
Capability
6/9/2015 UND Lunabotics 13
13. • Mine to a Depth of
30 cm
• Mine minimum of
100 kg of Regolith in
10 Minutes
• Utilize Light-Weight
Materials
6/9/2015 UND Lunabotics 14
14. • Belt System
– Solid Belt
– Ladder Structure
• Scoop Design
– One Large Scoop
– 1-2-1 Pattern
• Folding of System
– Motor Controlled
– Spring Loaded
6/9/2015 UND Lunabotics 15
15. • Added Onboard
Storage
– Increased Regolith
Collection Ability
• Dumping of
Regolith
– Cam System
– Conveyer System
6/9/2015 UND Lunabotics 16
18. • Density is 1.74 g/cm3
– Aluminum is 2.7 g/cm3
• Strength was Lower,
but Specific Strength
is Higher
• Potential
Complications in
Machining &
Welding
6/9/2015 UND Lunabotics 19
19. • Carbon Fiber
Design
• Small Cup Size
Allows for Better
Digging
– Reduced Cutting
Edge
6/9/2015 UND Lunabotics 20
22. • Camera Calibration
• One Frame Needed for
Detection
• Detection Phase
– SIFT (Scale Invariant
Feature Transform)
Feature Detectors
– SIFT Descriptors
• Pose Estimation
– With descriptors we
compare to our 3D
model of the image to
find 3D position of
6/9/2015 UND Lunabotics 23
24. • Testing
– Scoop Shape
– Force of Mining
Motion
– Necessary Motor
Size
• Conclusions
– Shortened Scoop
6/9/2015 UND Lunabotics 25
25. • Blocked Rotor Testing
– Peak at 20A
– Back EMF
• Traction Control Testing
– Response time shown
under no load condition
6/9/2015 UND Lunabotics 26
Top: Results from Block Rotor Testing
Bottom: Actual vs. Desired RPMs in Traction Control
Testing
27. • Modular System
• Attained Mining Depth of 50 cm
• Increased Regolith Storage
• Autonomous Capable System
6/9/2015 28UND Lunabotics
28. Thanks to Our
Advisors:
– Dr. Jeremiah Neubert
– Dr. Surojit Gupta
– Dr. Naima Kaabouch
Credits:
1. Lowe, David G. “Distinctive image
features from scale-invariant
keypoints.” International Journal of
Computer Vision 60.2 (2004):91-110
Thanks Again to Our
Sponsors:
6/9/2015 UND Lunabotics 29
32. Start
Initialize
Start
Threads
Localization
Thread
Obstacle
Detection
Thread
Main
Thread
Micro-
Processors
Thread
runloop runloop runloop runloop
Find Ground
Plane using
RANSAC
Find Obstacles
Send Obstacle
Information to
Main Thread
Grab Information
from the Other
Threads
Run Kalman Filter
Send PWM data
Receive Data
from Micro-
Processors
Send Micro-
Processor Data
to Main Thread
Send PWM data
to Micro-
Processor Thread
Find Features,
Compute
Descriptors
Compare
Desciptors with
3D Point Cloud
Calculate Pose
and Send to
Main Thread
System Architecture