Transcript Sample Presentation Title

Autonomous Tracking
Robot
Andy Duong
Chris Gurley
Nate Klein
Wink Barnes
Georgia Institute of Technology
School of Electrical and Computer Engineering
ECE4007, L04
April 24th, 2009
Project Overview
• Autonomous tank with the ability to
– Locate target
– Track and follow target
– Aim and fire projectile
• 1/16 scale proof-of-feasibility
• Save lives by removing humans from the front line of battle
• Development costs under $4,000
2
Vehicle Design
3
Design Objectives
• Locate a target:
– Detect presence of target using IR sensors
– Rotate to face target
• Track and follow:
– Approach target to within 1.5m ± .25m
– Correct course with color tracking software
– Ultrasonic sensor to determine range
• Aim and fire projectile:
– Aim turret using color tracking software
– Fire projectile at target with a 6° cone of accuracy
4
Project Illustration: Target Acquisition
5
Project Illustration: Rotation and Approach
6
Hardware Overview
7
Hardware Design
8
Relay Circuit
9
Picture of Relay Circuits
10
Hardware: Sensors
Ultrasonic Sensor
Detection Range
IR Sensor
IR Spot Sensor
30cm
30cm
5m
5m
Horizontal Detection
100°
38°
Vertical Detection
82°
22°
Movement Detection
Detection Distance (max)
11
0.1524m (6 inches) - 6.45m (254 inches)
Hardware Pictures
12
Software Components
• Windows CE Embedded OS
• Software in C++
• Phidgets 8/8/8 Board to read sensor inputs, send digital
output signals
• “phidgets21” library for function calls
– Analog methods return value between 0-1000
– IR sensor value >380 or <620 indicates target motion
– IR Sensors steady state around 505 (~2.5V)
– Spot sensor value >475 or <535 indicates aligned target
13
Software Algorithm Overview
14
Color Tracking Algorithm
• Edit webcam test software for color tracking algorithm
• Compare pixels and create groups of horizontal pixels of the
target color (red)
– A target is red if: R > G + B where R, G, and B are the
red, green, and blue values in a pixel
• Compare neighboring horizontal pixel groups to form blobs
• Average the indexes of the largest blob and turn the tank
based on the blob’s average offset from the center pixel
15
Problems
• Sensor/Webcam fluctuations reduce accuracy
– Over 1m, moving target caused sporadic values
– Webcam auto-corrects color, brightness, contrast
• Relative color tracking algorithm
• More power needed than expected
– Digital outputs unable to provide sufficient current
– USB does not provide enough power
• Spot sensor value hardcoded for specific location
16
Limitations
• Tracking target
– Accumulation of error in turret rotation
– Ground slippage
– Vertical alignment not viable
17
Budget and Cost Analysis
Item
RC German Tiger Tank
AMN23111 IR Motion Sensor
AMN21111 IR Spot Sensor
HB-401 H-Bridge
Qty
Item (Borrowed)
Qty
Total Cost
1
4
1
1
$79
$144
$12
$50
Est. Cost
QuickcamPro 9000 Webcam
1
$70
Phidgets Ultrasonic Sensor
1
$30
Wiring/Mounts/Transistors
1
$5
-Total Amount Spent: $285
-Total Estimated Parts Cost: $390
-Total Estimated Cost (1 Unit, Including Labor): $3260
-Total Proposed Selling Price (Based on Mass Production): $3600
18
Results
Goals
Proposed
Actual
Detect Target
YES
YES
Approach Target Distance
1.5m ± .25m
1.15m
Relocate Target if Lost
YES
YES
Firing Accuracy
6°Cone
5° Cone
19
Demonstration and Acceptance Testing
• Actions:
1. Detecting and rotating to face target
2. Approaching and tracking target
3. Aligning turret and firing projectile
Testing Stages Description
Independent testing of each action
One
Two
Testing of consecutive actions
Three
Testing of all actions consecutively
20
Future Improvements
21
Suggested Improvement
Benefit
Add encoder to turret
Higher accuracy alignment to target
Use multiple ultrasonic sensors
More accurate distance readings
Attain a higher quality camera
Fewer false movements during color
tracking
Include push buttons to start/stop
algorithm
Provide more control when
testing/demonstrating
Hide hardware inside tank
Aesthetics; Less chance of wires
interfering with tank mechanics
22