Transcript SoccerRobotsFinal - University of Adelaide

2005 Level IV Design Project
SOCCER ROBOTS
Michael Hill
Nicholas Jones
Michael Shanahan
Supervisor: Dr Frank Wörnle
RoboCup
•
International robotic
soccer competition
•
Promotes studies in:
•
•
•
Automation
Artificial Intelligence
Strategic
Programming
Goals & Objectives
•
Research & develop a design meeting
RoboCup small-size league entry criteria
•
Manufacturing a team of three such robots
•
Development & implement a vision system to enable
multi-agent coordination
Goals & Objectives
•
Design & implement a communications system to
facilitate automation & strategic game-play
•
Comply with the AU$250.00 budget
•
Provide a solid foundation for future project teams to
build upon & achieve success
Design Overview
•
•
•
•
•
•
Mechanical Design
Vision System
AI Software
Communications System
Current & Future Work
Conclusion
Mechanical Design
•
Drive System
•
Kicker Device
•
Power
•
Chassis
Mechanical Design
•
Drive System
•
Two Wheel Differential
•
Omnidirectional
University of Auckland, 2005
Weber, 2004
Team Lucky Star, 2004
Mechanical Design
•
Drive System
Final Design: A differential drive system
using M42SP-5 stepper motors.
•
•
•
Cost effective
Readily available
Satisfy torque and speed requirements
Mechanical Design
•
Drive System – Motor Selection
Performance Specifications
• Acceleration 1 m/s2
• Velocity 1 m/s
Parameters
• Robot Mass: 1.9 kg
• Wheel Radius: 30 mm
• Step Size: 7.5o
• Motor Speed: 365 pps
MGEN, 2005
Required Torque: 43 mN.m
Maximum Speed: 1.4 m/s
Mechanical Design
•
Kicker Device
•
Spring Powered
•
Pneumatics
•
Solenoid
Weber, 2004
Mechanical Design
•
Kicker Device
Final Design: Pull-type
solenoid with lever
•
Cost effective
•
Simple to build
Mechanical Design
•
Power
Powered Devices:
•
MiniDragon+ development board (5V)
•
Stepper Motors and Drivers (12V)
•
Solenoid (12V)
•
RF Receiver (5V)
Power Source:
•
Ten 1.2V rechargeable AA batteries in
series
Mechanical Design
•
•
Chassis
Constructed from
3 mm aluminium
Mattner, 2005
•
Modular design to
aid in modification
of subsystems
Vision System
• Equipment
Fire Wire Camera (15 FPS)
Frame
Vision Software
The Imaging Source, 2005
Mattner, 2005
Vision System
Colour Identification System
Pink = Centroid r1
Purple = Direction r1
Yellow = Centroid r2
Blue = Direction r2
Orange = Ball
Vision System
Object Tracking
Vision System

Takes an image

Returns it to the vision software

Locates the positions

Returns them to the AI System
AI Software
•
Controls all robotic movement & kicking
•
MATLAB code & Simulink block diagrams
converted to C-code
•
Two phases of operation
1.
2.
Ball not in possession
Ball in possession
AI Software
•
Ball Not In Possession

Assumes rate of
acceleration is constant
over small sample time

Loop performed every
five samples
AI Software
•
Ball in possession

Constant maxKick defines
furthest the ball can be
projected by kicker device

Ball is kicked once
dist2Goal is less than
maxKick

Similar algorithm for passing
AI Software
•
Problem: Data reception too slow for
effective closed loop navigation
•
Solution: AI provides several instructions
per iteration
•
Development of game-play strategy limited
due to lack of opposition
Communications
System
•
Robot instructions are sent in 8-byte telegrams via
radio transceivers
•
Signals are broadcast to all three robots on same
frequency
•
Data Transfer Rate = 12.5 instructions per second
Communications
System
•
Bytes 1-4: Telegram Information
Contains information about the telegram itself, including target
robot
•
Bytes 5 & 6: Motor Instructions
Signed Integer varying between -127 to +127 representing
speed and direction for the left and right motors, respectively
Communications
System
•
Byte 7: Solenoid Instructions
Set to 1 for energised, 0 for de-energised
•
Byte 8: Checksum
Sum of transmitted bits, which must be equal to the sum of
received bits
Current Work
•
Successful in achieving:
•
Effective & cost efficient design
•
Vision System provides object tracking
•
AI Software with predictive motion nearing
completion
•
Communications system allows remote interaction
Future Work
•
Group Aims:
•
Functional team of three robots
•
Perform standard soccer actions
e.g. pass, shoot & dribble
•
Ball trajectory prediction software
•
Communications system supports continuous
robot coordination
Future Work
•
Recommended Development:
•
Manufacture opposition team & develop defensive
AI software
•
Improve vision system data rates to facilitate faster
game-play
•
Incorporate strategic game-play
Conclusion
•
Project budget was the most defining
constraint
•
Goals set were inherently optimistic
•
Were able to achieve majority of all
fundamental hopes for project
Acknowledgements
We would like to thank the following people
Dr Frank Wörnle, our supervisor
Bill, Richard and Steve from the mechanical workshop
Silvio and Derek from the electronics workshop
Carnegie Mellon University for the vision drivers
Yasutake and Taiki from the 2005 Osaka University Team
Questions???