Presentation is loading. Please wait.

Presentation is loading. Please wait.

Mechanical Department RoboCup 19 September, 2005 The University of Adelaide Copyright © 2005Slide Number 1 David Jackson Nathan Simmonds Robert Stewart.

Similar presentations


Presentation on theme: "Mechanical Department RoboCup 19 September, 2005 The University of Adelaide Copyright © 2005Slide Number 1 David Jackson Nathan Simmonds Robert Stewart."— Presentation transcript:

1 Mechanical Department RoboCup 19 September, 2005 The University of Adelaide Copyright © 2005Slide Number 1 David Jackson Nathan Simmonds Robert Stewart Supervisor: Dr. Frank Wornle

2 Mechanical Department RoboCup 19 September, 2005 The University of Adelaide Copyright © 2005Slide Number 2 RoboCup Organisation  Research initiative in the fields of robotics and artificial intelligence  Soccer environment includes friendly, neutral and adversarial agents  RoboCup aim: To beat the human world champions by 2050 (www.robocup.org)

3 Mechanical Department RoboCup 19 September, 2005 The University of Adelaide Copyright © 2005Slide Number 3 Project Aim  To develop skills such as ball collection, obstacle avoidance and kicking Intelligence System  Decision making algorithms  Path Planning Vision Processing Camera Motors Controller Kicker Camera

4 Mechanical Department RoboCup 19 September, 2005 The University of Adelaide Copyright © 2005Slide Number 4 Vision System A xaxa Robot Body B xbxb x Sensor y Robot Body C xcxc

5 Mechanical Department RoboCup 19 September, 2005 The University of Adelaide Copyright © 2005Slide Number 5 Constant Curvature Discontinuity  Image continuity is not preserved  Vision of the constant curvature section overlaps the isometric  This error was caused by an assumption made about the camera focal length  Maximum range: 8.5m

6 Mechanical Department RoboCup 19 September, 2005 The University of Adelaide Copyright © 2005Slide Number 6 Vision Calibration  CMVision – Classifies pixels as a YUV colour  CMVision returns pixel information of colour regions  Obtaining the physical distance of an object requires mirror calibration  Obtain the pixel distance and map this to the measured physical distance

7 Mechanical Department RoboCup 19 September, 2005 The University of Adelaide Copyright © 2005Slide Number 7 Pixel-Distance Transformation

8 Mechanical Department RoboCup 19 September, 2005 The University of Adelaide Copyright © 2005Slide Number 8 Localisation  Locate blue and green goalposts  The centre of the goal-line is the global origin  The robot orientation is the slope of the goal-line  Using the global position and orientation of the robot, the global position of an obstacle may be obtained from its position relative to the robot

9 Mechanical Department RoboCup 19 September, 2005 The University of Adelaide Copyright © 2005Slide Number 9 Localisation  The projection of the ball causes the red region to be returned  The blue region represents the coordinates of interest  Using a compensation function, the ball may be located with an error of less than 8cm

10 Mechanical Department RoboCup 19 September, 2005 The University of Adelaide Copyright © 2005Slide Number 10 Intelligence System  Controls and maintains links between vision system and controller  Responsible for receiving and transmitting data from the subsystems  Decides robots next location and/or action  Uses path planner to obtain intermediate goals  Determines if intermediate goals have been reached through a combination of vision updates and encoder odometry  Updates camera at timed intervals  Transmits commands to controller on demand

11 Mechanical Department RoboCup 19 September, 2005 The University of Adelaide Copyright © 2005Slide Number 11 Decision making process  Single player state  Retrieve ball  Goal is set behind ball on line from goal  Shoot for goal  In case of obstacles between the ball and goal, pass towards danger zones and/or team mates

12 Mechanical Department RoboCup 19 September, 2005 The University of Adelaide Copyright © 2005Slide Number 12 Heuristic  Possession of ball  Ball in view  Robot orientation  Angle to goal  Distance to goal  Obstacles in path

13 Mechanical Department RoboCup 19 September, 2005 The University of Adelaide Copyright © 2005Slide Number 13 Future Versions  Multiplayer state  Wireless communication between teammates  More complex decisions eg Pass/Shoot/Dribble  Player modes –Defender –Goal keeper –Attacker –Midfielder  Players controlling field zones  Game modes –Man on Man –Offensive –Defensive –Long range shots –Space seeking –Time wasting –Aggressive

14 Mechanical Department RoboCup 19 September, 2005 The University of Adelaide Copyright © 2005Slide Number 14 Path Planning  Receives position and geometric information about every object in the global frame  This module must produce a route that: 1.Avoids all obstacles 2.Accurately and reliably finds the destination 3.Minimises the path distance 4.Is computed in minimal time

15 Mechanical Department RoboCup 19 September, 2005 The University of Adelaide Copyright © 2005Slide Number 15 Geometric Algorithm  Uses geometry to find the optimal path using via-points  The red circle is the obstacle, the blue circle is the minimum allowable distance from the obstacle and the crosses are via-points Start Goal VP 1 VP 2 o g VP 1 VP 2 Start Goal

16 Mechanical Department RoboCup 19 September, 2005 The University of Adelaide Copyright © 2005Slide Number 16 Geometric Algorithm

17 Mechanical Department RoboCup 19 September, 2005 The University of Adelaide Copyright © 2005Slide Number 17 Evaluation  Can solve complex paths  Near optimal path generation  Not infallible  Computation Time < 150ms  Does not yet account for moving obstacles

18 Mechanical Department RoboCup 19 September, 2005 The University of Adelaide Copyright © 2005Slide Number 18 Navigation System  Omnidirectional  Two configurations  Allows rotation in conjunction with translation  Encoder feedback 120 degree configuration 150 degree configuration (Weber, 2004)

19 Mechanical Department RoboCup 19 September, 2005 The University of Adelaide Copyright © 2005Slide Number 19 Controller  Designed on the DRAGON12 Development Board  Commands received from laptop via serial port (RS-232)  Responsible for controlling the three motors to move at specified velocities through the implementation of three separate PID feedback loops  Responsible for controlling the inlet/outlet valves of the kicker and its clutch release (http://www.evbplus.com/dragon12.html)

20 Mechanical Department RoboCup 19 September, 2005 The University of Adelaide Copyright © 2005Slide Number 20 PWM for Motor3 PWM for Motor2 PWM for Motor1 Encoder feedback for Motor2 Encoder feedback for Motor3 ControllerPlant + - Motor Speed PID Control Controller  Responsible for generating PWM signals and Direction bits to motor driver  Responsible for reading Motor Encoder input  Responsible for PID feedback loop using Encoder  Responsible for controlling the inlet/outlet valves of the kicker and its clutch release Air Inlet Air Outlet Clutch

21 Mechanical Department RoboCup 19 September, 2005 The University of Adelaide Copyright © 2005Slide Number 21 Controller Encoder Feedback: 2 16-Bit Pulse Accumulators 1 Rising-Edge Accumulator Parse Command SCI_ISR() - Command over serial port Raise flag Heartbeat overflow_ISR() – Timer overflow dir_ISR()- Counter overflow Increment Counter Lower flag If flag raised main Kicker Control PID Feedback Examples Velocity Command “V 1024 -512 0 \n” Kicker Rod Pullback Command “T 3\n” Release Clutch “K \n” Discrete Time Interval for Information Processing Information Processing -Calculate the period of the encoder pulses for PID feedback loop -Kicker Control - inlet/outlet of air to muscle -Heartbeat – if no signal been received by controller, switch off motors Calculate period of encoders pulses

22 Mechanical Department RoboCup 19 September, 2005 The University of Adelaide Copyright © 2005Slide Number 22 Kicker System  Designed in 2004 by George Osborne and Christian Weber  Based on a combination of a tension spring and an air muscle using a spring clutch mechanism. (Weber, 2004)

23 Mechanical Department RoboCup 19 September, 2005 The University of Adelaide Copyright © 2005Slide Number 23 Kicker System Air Inlet (Weber, 2004)

24 Mechanical Department RoboCup 19 September, 2005 The University of Adelaide Copyright © 2005Slide Number 24 Summary  The project aims to integrate these hardware units into a reliable and intelligent software system capable of deriving an optimal solution to the tasks of avoiding obstacles, and of finding, collecting and kicking the ball. Current goals to achieve project aims  Integration of individual software and hardware components  Testing and tuning for obstacle avoidance  Collecting and kicking the ball during navigation

25 Mechanical Department RoboCup 19 September, 2005 The University of Adelaide Copyright © 2005Slide Number 25 Questions?

26 Mechanical Department RoboCup 19 September, 2005 The University of Adelaide Copyright © 2005Slide Number 26 (Weber, 2004) Slides for questions: Kicker Shot

27 Mechanical Department RoboCup 19 September, 2005 The University of Adelaide Copyright © 2005Slide Number 27


Download ppt "Mechanical Department RoboCup 19 September, 2005 The University of Adelaide Copyright © 2005Slide Number 1 David Jackson Nathan Simmonds Robert Stewart."

Similar presentations


Ads by Google