2011 Simulated Car Racing Championship @ GECCO-2011

2,750 views

Published on

Summary f the second leg of the 2011 Simulated Car Racing Championship held during GECCO-2011

More information at

http://cig.dei.polimi.it
http://groups.google.com/group/racingcompetition
http://sourceforge.net/projects/cig/

Organizers
Daniele Loiacono, Politecnico di Milano
Luigi Cardamone, Politecnico di Milano
Martin Butz, University of Würzburg
Pier Luca Lanzi, Politecnico di Milano

Published in: Technology, Sports, Automotive
0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total views
2,750
On SlideShare
0
From Embeds
0
Number of Embeds
423
Actions
Shares
0
Downloads
61
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide

2011 Simulated Car Racing Championship @ GECCO-2011

  1. 1. The 2011 Simulated Car Racing Championship @ GECCO-2011 Daniele Loiacono, Luigi Cardamone, Martin V. Butz, and Pier Luca Lanzi2011 Simulated Car Racing Championship @ GECCO-2011
  2. 2. 2 2011 Simulated Car Racing Championship 9 races during 3 conferences Develop a driver for TORCS (hand-coded, learned, evolved, …) Drivers will be awarded based on their score in each conference competition At the end, the team with highest overall score wins the championship2011 Simulated Car Racing Championship @ GECCO-2011
  3. 3. Roadmap of the 2011 Championship 1.  EVO*-2011, Torino (Italy) April 27-29, 2011 2.  ACM GECCO-2011, Dublin (Ireland) July 12-16, 2011 3.  IEEE CIG-2011, Seoul (South Korea) August 31 September 3, 20112011 Simulated Car Racing Championship @ GECCO-2011
  4. 4. Motivations q  Proposing a relevant game-based competition " more representative of commercial games AI " more similar to a real-world problem q  Proposing a funny and exciting competition " you can see and play with the entries of this competition " human players can interact with AI " a lot of programmed AI available for comparison q  Proposing a challenging competition " computationally expensive " real-time " on-line learning " noisy sensors2011 Simulated Car Racing Championship @ GECCO-2011
  5. 5. The Open RacingCar Simulator
  6. 6. The Open Racing Car Simulator & the Competition Software TORCS TORCS PATCH BOT BOT BOT SBOT SBOT SBOT q  The competition server UDP UDP UDP q  Separates the bots from TORCS q  Build a well-defined sensor model q  Works in real-time BOT BOT BOT2011 Simulated Car Racing Championship @ GECCO-2011
  7. 7. Sensors and actuators q  Rangefinders for edges on the track and opponents (with noise) q  Speed, RPM, fuel, damage, angle with track, distance race, position on track, etc. q  Six effectors: steering wheel [-1,+1], gas pedal [0, +1], brake pedal [0,+1], gearbox {-1,0,1,2,3,4,5,6}, clutch [0,+1], focus direction2011 Simulated Car Racing Championship @ GECCO-2011
  8. 8. 8 What is the structure of a race? Three stages: warm up, qualifiers, actual race During warm-up, each driver can explore the track and learn something useful During qualifiers, each driver races alone against the clock (the best 8 drivers move to the race) During the race all the drivers race together2011 Simulated Car Racing Championship @ GECCO-2011
  9. 9. Competitors
  10. 10. The competitors q  Four entries in the first leg: " Mr. Racer, TU Dortmund " Mariscal & Fernández, Málaga University " Ready2Win, Slovak University of Technology - FIIT " Powaah, Blekinge Institute of Technology q  Two more entries in the second leg: " Ho Duc Thang, University of Nottingham " CRABCAR, Norwegian University of Science and Technology (NTNU) q  Three reference entries from the past championship: " AUTOPIA, Madrid and Granada " COBOSTAR (Lönneker & Butz, University of Würzburg) " POLIMI (Cardamone, Politecnico di Milano)2011 Simulated Car Racing Championship @ GECCO-2011
  11. 11. Mr Racer Jan Quadflieg, Tim Delbruegger and Mike Preuss TU Dortmund2011 Simulated Car Racing Championship @ GECCO-2011
  12. 12. Mr. Racer 2011 q  13 Variables learned offline with the CMA-ES q  Warmup used to learn a model of the track q  Noise Handling: " Low pass filter on sensor values " 2 Regression polynoms are fitted to each side " Resample the polygons q  Overtaking " Computes on the fly a recommended speed and an overtaking line " Recommendation are provided to a planing module2011 Simulated Car Racing Championship @ GECCO-2011
  13. 13. 2011 Simulated Car Racing Championship @ GECCO-2011
  14. 14. Mariscal & Fernández q  General approach: expert systems improved by multi- objectives evolutionary computation techniques " Maximize distance. " Minimize damage q  Steering based on a simple rule follow the track sensor with biggest distance. q  Overtaking based on simple rules and expert systems2011 Simulated Car Racing Championship @ GECCO-2011
  15. 15. 2011 Simulated Car Racing Championship @ GECCO-2011
  16. 16. Ready2Win q  Modular architecture: " Driving module: •  The speed is computed using the maximum braking distance " Overtaking module " Recovery module " ABS module q  Track learning during the warm-up: " First lap, driven slow, to identify turns (start, end, entry position, curvature) and learn the track model " Other laps, speed adaptation: •  Increasing according to lateral speed •  Reducing when the car goes off-road2011 Simulated Car Racing Championship @ GECCO-2011
  17. 17. Powaah Tim Uusitalo Blekinge Institute of Technology Supervisor: Stefan Johanson2011 Simulated Car Racing Championship @ GECCO-2011
  18. 18. Powaah q  Based on Artificial Potential Field q  Every sensor in the span (-40, +40) has a charge q  The charge is determined by how far the sensor measures the track q  Coulomb’s law is used to calculate potentials in the field Pot = (C1 * C2 ) / d2 q  The sensor with the highest potential is used for computing how much steering is needed2011 Simulated Car Racing Championship @ GECCO-2011
  19. 19. Ho Duc Thang University of Nottingham2011 Simulated Car Racing Championship @ GECCO-2011
  20. 20. Ho Duc Thang q  Hybrid fuzzy controller with a module to learn the model track as the car drives, a module to perform some simple plannings based on the modelled segments of the track and a fuzzy controller which actually drives the car. q  All the modelling, planning and driving are done during the race, the warm-up stage is not exploited q  No opponent handling2011 Simulated Car Racing Championship @ GECCO-2011
  21. 21. CRABCAR By Snorre Corneliussenand Magnus Westergaard, NTNU
  22. 22. CRABCAR q  Turns classification based on sensors during warmup and saved in the track model q  Different evolved strategies to go through the turns q  Strategies are re-evaluated during stages, adjusted and save in the track model q  Simple opponent avoidance based on keeping safe distance2011 Simulated Car Racing Championship @ GECCO-2011
  23. 23. AUTOPIA Industrial Computer Science Department. Centro de Automática y Robótica Consejo Superior de Investigaciones Científicas Madrid, Spain Contact:E. Onieva (enrique.onieva@car.upm-csic.es)2011 Simulated Car Racing Championship @ GECCO-2011
  24. 24. AUTOPIA q  Fuzzy Architecture based on three basic modules for gear, steering and speed control " optimized with a genetic algorithm q  Learning in the warm-up stage: " Maintain a vector with as many real values as tracklength in meters. " Vector initialized to 1.0 " If the vehicle goes out of the track or suffers damage then multiply vector positions from 250 meters before the current position by 0.95. q  During the race the vector is multiplied by F to make the driver more cautious in function of the damage: " F=1-0.02*round(damage/1000)2011 Simulated Car Racing Championship @ GECCO-2011
  25. 25. COBOSTAR Thies Lönneker and Martin V. Butz University of Würzburg http://www.coboslab.psychologie.uni-wuerzburg.de2011 Simulated Car Racing Championship @ GECCO-2011
  26. 26. COBOSTAR q  One of the best controller of the 2009 Championship q  Parameterized controller optimized with CMA-ES q  Dynamically saves crash points2011 Simulated Car Racing Championship @ GECCO-2011
  27. 27. Cardamone Luigi Cardamone Politecnico di Milano2011 Simulated Car Racing Championship @ GECCO-2011
  28. 28. Luigi Cardamone q  Winner of the CIG-2008 Car Simulated Competition q  The controller is based on a neural network evolved with NEAT q  Very simple policy for overtaking2011 Simulated Car Racing Championship @ GECCO-2011
  29. 29. Qualifying
  30. 30. Scoring process: Warm-up & Qualifying q  Scoring process involves three tracks: " Cold Peak " City Track " Desert Storm q  The tracks are not distributed with TORCS: " Generated through interactive evolutionary computation " The competitors cannot know the tracks q  Each controller raced for 100000 game ticks in the warm-up stage and then its performance is computed in the qualifying stage as the distance covered within 10000 game ticks2011 Simulated Car Racing Championship @ GECCO-2011
  31. 31. Dirt-1City Track
  32. 32. Qualifying: City Track CRABCAR 8392.34 Cardamone 8980.1 COBOSTAR 9439.67 Powaah 10825.3 Mariscal & Fernandez 10948.8 Ho Duc Thang 11177.2 Mr. Racer 11420 Ready2Win 11424.4 AUTOPIA 11565.4 0 2000 4000 6000 8000 10000 12000 140002011 Simulated Car Racing Championship @ GECCO-2011
  33. 33. Dirt-1Cold Peak
  34. 34. Qualifying: Cold Peak Cardamone 7727.57 CRABCAR 9155.7COBOSTAR 9166.16 Mariscal & 9270.04 Fernandez Ho Duc 9286.447266 Thang Ready2Win 9924.31 Powaah 10233.1 Autopia 10347.5 Mr. Racer 10864.9 0 2000 4000 6000 8000 10000 120002011 Simulated Car Racing Championship @ GECCO-2011
  35. 35. Dirt-1Desert Storm
  36. 36. Qualifying: Desert Storm Cardamone 5575.03 COBOSTAR 9082.14 CRABCAR 9297.25 Ho Duc 9954.833984 Thang Ready2Win 10468.4 Mariscal & 10603.4 Fernandez Powaah 10740.3 Mr. Racer 10774.8 AUTOPIA 11159 0 2000 4000 6000 8000 10000 120002011 Simulated Car Racing Championship @ GECCO-2011
  37. 37. Qualifying summary Driver CityTrack ColdPeak DesertStorm Total AUTOPIA 10 8 10 28 Mr. Racer 6 10 8 24 Ready2Win 8 5 4 17 Powaah 3 6 6 15 Mariscal & Fernandez 4 3 5 12 Ho Duc Thang 5 4 3 12 COBOSTAR 2 2 1 5 CRABCAR 0 1 2 3 Cardamone 1 0 0 12011 Simulated Car Racing Championship @ GECCO-2011
  38. 38. Qualifying summary Driver CityTrack ColdPeak DesertStorm Total AUTOPIA 10 8 10 28 Mr. Racer 6 10 8 24 Ready2Win 8 5 4 17 Powaah 3 6 6 15 Mariscal & Fernandez 4 3 5 12 Ho Duc Thang 5 4 3 12 COBOSTAR 2 2 1 5 CRABCAR 0 1 2 3 Cardamone 1 0 0 1 Evolutionary Approaches2011 Simulated Car Racing Championship @ GECCO-2011
  39. 39. Races
  40. 40. 40 Three Tracks For each track we run 8 races with random starting grids Each race is scored using the F1 point system (10 to first, 8 to second, 6 to third, …) Two points to the controller with lesser damage Two points for the fastest lap of the race2011 Simulated Car Racing Championship @ GECCO-2011
  41. 41. Final Results Competitor CityTrack ColdPeak DesertStorm Total AUTOPIA 12 10 12 34 Mr. Racer 6.5 8.5 6 21 Ready2Win 5 6 8 19 Mariscal & Fernandez 5.5 3.5 5.5 14.5 Ho Duc Thang 5.5 5.5 3 14 COBOSTAR 3 3.5 4 10.5 CRABCAR 2 2 3 7 Powaah 2 1.5 2 5.52011 Simulated Car Racing Championship @ GECCO-2011
  42. 42. What  about  the  race  results?     Race  Start  is  crucial!     AUTOPIA  s9ll  the  best  controller…    Qualifying  and  final  results  very  similar  (with   few  excep9ons)      
  43. 43. Championship’s Standings Competitor EVO* GECCO TOTAL AUTOPIA 34 34 68 Mr. Racer 22 21 43 Ready2Win 10 19 29 Mariscal & Fernandez 20 14.5 34.5 Ho Duc Thang - 14 14 COBOSTAR 17 10.5 27.5 CRABCAR - 7 7 Powaah 10 5.5 15.5 Cardamone 12 - 12 Next leg @ CIG-2011 in Seoul2011 Simulated Car Racing Championship @ GECCO-2011

×