• Share
  • Email
  • Embed
  • Like
  • Save
  • Private Content
Demolition Derby 2012 - GECCO Competition Report
 

Demolition Derby 2012 - GECCO Competition Report

on

  • 562 views

Slides about the competition including the results and the winner.

Slides about the competition including the results and the winner.

Statistics

Views

Total Views
562
Views on SlideShare
562
Embed Views
0

Actions

Likes
0
Downloads
1
Comments
0

0 Embeds 0

No embeds

Accessibility

Categories

Upload Details

Uploaded via as Microsoft PowerPoint

Usage Rights

© All Rights Reserved

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
  • Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
    Processing…
Post Comment
Edit your comment
  • Optimization of avoidance is as important as is the optimization of hitting. Co-optimization is possible. Point out that various optimization methods can be applied – policy-gradient algorithms can be done with CMA-ES for example. Strategy components can be optimized individually or in parallel.
  • Last year: Both competitors did NOT consider opponent AVOIDANCE. However, chasing and crashing was attempted to be optimized. Result was, however, rather unclear outcomes with 8 competitors.

Demolition Derby 2012 - GECCO Competition Report Demolition Derby 2012 - GECCO Competition Report Presentation Transcript

  • Institute of Computer Science Chair of Cognitive ModelingDemolition Derby 2012Based on TORCS: The Open Racing Car Simulator07/2012, Martin V. Butz
  • Demolition Derby 2012 http://cm.inf.uni-tuebingen.de/competitions• Organized by: - Martin V. Butz, University of Tübingen, Germany• Supported by: - Andreas Alin, University of Tübingen, Germany - Dennis Schwartz, University of Tübingen, Germany• And with previous help by: - Matthias J. Linhardt, University of Bamberg, Germany - Daniele Loiacono, Politecnico di Milano, Italy - Luigi Cardamone, Politecnico di Milano, Italy - Pier Luca Lanzi, Politecnico di Milano, Italy
  • Demolition Derby: Purpose• Optimize opponent interactions - Avoid being hit – run away when necessary - Try to hit others at the right moment.• Enables (co-)optimization of interaction behavior. behavior - Fitness may be based on damage caused to other cars. - Co-development of two or more competitors is possible (possibly with different approaches). - Can do policy-gradient-based optimization• Various strategy components are relevant - Avoidance optimization - Chasing others optimization - Forwards & backwards steering control - Opponent monitoring - Meta-strategies3
  • Goal & Setup• Goal: Wreck all opponent cars by crashing into them without getting wrecked yourself.• Setup: Local sensor information as in the Simulated Car Racing Competition.• Sensors: - Simulated distances sensors (noiseless).  Surrounding 36 opponent sensors with a range of 300m.  19 track sensors with a rang of 200m. - Other sensors  Current damage of own car.  Damage produced on other cars.  Status of car (speed, wheels, gear…).  Relative position on track. - Damage model:  Cars do not take any damage when colliding with walls.  Cars do not take any damage in the front when colliding with each other.  Cars only take damage when their rear is hit by another car.4
  • Winner Determination• Arena: Large circular track (surface: asphalt; length: 640m, width: 90m) Arena• Qualifying - 1-vs-1 matches evaluating all against all (winner = 1 point = less damage) - Eight best controllers qualify for the final showdown.• Final demolition derby matches: - The best eight controllers fight each other. - Ten matches are played. - Car that wins most often wins the competition. - Alternative scoring with rank-based points is also considered. 5
  • Additional Goodies for a Quick Start• Basic controller clients for Java and C++, to easily add additional functionality.• COBOSTAR client in Java - With opponent monitor that tracks opponents over time. - With simple crashing strategy that targets closest car in range.• Evolvable client setup that - receives caused damage signal, - applies CMA evolution strategy-based optimization, - runs continuously with or (even faster) without visualization for as many generations as desired.
  • Last Years Entries• Base Client - Dep. of Computer Science - University of Würzburg, Germany• DemoStar - Thies Lönneker, Dep. of Computer Science - University of Würzburg, Germany• Spartiat - Zygmunt Horodyski, Piłsudskiego 39/1 - 66-530 Drezdenko, Poland
  • This Year’s NEW Entries• JustDetermined - Brian J Tellier - University of Southern Maine (USM), Portland, ME, USA USM• KevinCar - Kevin Knowlton - University of Southern Maine (USM), Portland, ME, USA USM• SEALbot - Anderson Rocha Tavares & Gabriel de Oliveira Ramos & Renato de Pontes Pereira & Sérgio Montazzolli Silva & Ana L. C. Bazzan - Universidade Federal do Rio Grande do Sul (UFRGS), Porto Alegre, UFRGS Brasil• SloppyJalopy - David Gagne - University of Southern Maine (USM), Portland, ME, USA USM• WSI-Client 2012 - Wilhelm-Schickard-Institut of Computer Science - University of Tübingen, Germany8 © 2012 University of Tuebingen
  • Last Year: Demo Star Approach• Chase Opponent – try to crash• Avoidance not considered.• Challenges tackled: - Opponent Tracking  Use local distance sensors  Also use global information (track location) - DemoStar Agility:  CMA-ES-based optimization of turning. - DemoStar Recovery: Escape stuck situations  Stuck Face-to-face  Stuck at wall.9 © 2012 University of Tuebingen
  • This Year: Strategy Improvements• Simple approach from Uni Tübingen to increase the challenge: - Modified Base-Client that avoids attacks early in the game (“cowered car”).Approaches by new entries:• USM: Rule-based agents with higher-level rules and control strategies: - Conditions:  Ahead, Close Ahead, Behind, Advantage, Edge, Wounded, No Change, Duel, Turning - Actions:  Ram, Run, Bait, Get Clear, U-Turn, Circle Track• UFRGS: Fuzzy rule-based agent: - Four behavioral strategies:  Stuck, Evade, Attack (front, side), Look for Opponents - Scoring mechanism for selecting behavior and opponent - Fuzzy rules make decisions10 © 2012 University of Tuebingen
  • USM: Rule-based agents JustDetermined KevinCar SloppyJalopy11 © 2012 University of Tuebingen
  • USM Rule-Based Agents  Agent behaviors are determined using a rule-based approach  No learning for the entries here, but this approach is designed for EC learning  Conditions and actions are drawn from a discrete “vocabulary” of pre-designed options  Each rule is a condition-action vector  Conflict resolution:  Rule order here (but specificity is usually helpful)Gagne, Knowlton, Tellier, and Congdon, GECCO
  • USM Rule-Based Agents  Abstractions interface between game and rules  Low-level game sensors are abstracted to high-level rule conditions  High-level rule actions are translated to low-level game controls Game low-level sensors controls low-level details details high-level inputs outputs high-level abstractions Agent abstractions conditions actionsGagne, Knowlton, Tellier, and Congdon, GECCO
  • Conditions – Input Abstractions Condition TRUE when… Ahead Ahead An enemy is ahead and within 100 m Close Ahead Any enemy is ahead and within 20 m Behind An enemy is behind and within 100 m Behind Advantage Opponent has 2000 more damage than agent Edge Agent is near the track edge Wounded Agent has more than 7,000 damage Agent has been doing the same thing No Change for a long time Duel There is only one other opponent Turning Agent has started a U-turn, but has not finishedGagne, Knowlton, Algorithms Congdon, GECCOCongdon, Genetic Tellier, and and NonCoding DNA
  • Actions – Output Abstractions as Pictures Action Pictorial Description Ram Agent SMASH! Run Bait Get Clear U-Turn Circle TrackGagne, Knowlton, Tellier, and Congdon, GECCO
  • Actions – Output Abstractions in English Action Description Ram • Steer toward opponent • Slow down if necessary. • Otherwise, full acceleration. Run • Steer away from opponent • Circle the track • Full acceleration Bait • Circle the track • Speed limit 110 kph • When opponent is close, swerve Get Clear • Turn away from track edge • If very close to edge, back up U-Turn • Cut wheel hard left or right (coin flip) • Keep wheel cut for 100 steps Circle Track • Stay centered and in line with track axis • Speed limit 110 kphGagne, Knowlton, Tellier, and Congdon, GECCO
  • Sloppy Jalopy Entry – Rule Set Conditions ActionAhead Close Behin Advantage Edge Wounded No Change Duel Turnin Ahead d g F * * * T * * * Get Clear * * * F * F T * U-Turn * * * * F T * F Run * * * T F * * T Run * * T * F F * F Bait T * F * * * * F Ram T * * * * F F T Ram * * T F F * F T Bait * * * * * * F * Circle Track  ‘*’ Means any state satisfies this condition (Don’t Care).  Grayed-out conditions are ignored by this agent. Gagne, Knowlton, Tellier, and Congdon, GECCO
  • Sloppy Jalopy Behavior  SJ runs away when it’s wounded or winning a duel by a margin.  Rams only when there is nobody behind it.  Tries to score mainly by baiting opponents.  Is more timid in multiplayer, more aggressive in a duel.  Does a U-Turn if it’s been doing one thing for a while.  Circle track by default.Gagne, Knowlton, Tellier, and Congdon, GECCO
  • Crash and Segfault Entry – Rule Set Conditions ActionAhead Close Behin Advantage Edge Wounded No Change Duel Turnin Ahead d g F F * * T * * * * Get Clear * F T * F * F * F Run T F F T F * * * * Run T F F F * * * * F Ram F F F T F * * * * Circle Track * T * * * * * * * Ram * F T * F * T * * U-Turn F F F F F * F * F Circle Track F F F F F * T * * U-Turn F F * F F * * * T U-Turn ‘*’ Means any state satisfies this condition (Don’t Care). Grayed-out conditions are ignored by this agent. Gagne, Knowlton, Tellier, and Congdon, GECCO
  • Crash and Segfault Behavior  Will always attempt to ram if an opponent is close ahead.  Runs away if an opponent is behind it.  Runs away if it has a damage advantage in a duel.  Attempts to ram if there is an opponent ahead and it isnt running.  Will make a U-turn if it has gone a complete lap around the track while either running or circling.  If no other action is called for, will circle the track to try to find opponents.Gagne, Knowlton, Tellier, and Congdon, GECCO
  • JustDetermined Entry – Rule Set Conditions ActionAhead Close Behin Advantage Edge Wounded No Change Duel Turnin Ahead d g * * * * T * * * * Get Clear * * T * * * * * * U-Turn T * F * F * * * * Ram * * F T F * * * * Run F * F * F * * * * Circle Track  ‘*’ Means any state satisfies this condition (Don’t Care).  Grayed-out conditions are ignored by this agent. Gagne, Knowlton, Tellier, and Congdon, GECCO
  • JustDetermined Behavior  Basic wall detection to avoid walls  Opponent is its main focus when opponent is in front of the controller  If an opponent is behind the controller, it will turn around as fast as possible to hit the opponent  When no opponent is near, the controller circles the track to maintain speedGagne, Knowlton, Tellier, and Congdon, GECCO
  • Future Work  Further evaluation of agents against a wider variety of drivers.  This basic approach is designed as a step towards using EC on the rule sets.  In addition to evolving the rule sets, parameters such as “close” can benefit from EC to refine these values.Gagne, Knowlton, Tellier, and Congdon, GECCO
  • UFRGS: Fuzzy rule-based agent SEAL-Bot24 © 2012 University of Tuebingen
  • Results - Scoring• Preliminary 1 vs. 1 Matches - More matches won – participate in final showdown matches - Eight qualify – therefore, all qualify (eight entries in total)• Final showdown matches - 10 matches - Controller that wins the most matches wins the competition. - Alternative point scoring also reported.25 © 2012 University of Tuebingen
  • Preliminary 1 vs. 1 MatchesBaseClient WSI-Client 2012BaseClient Spartiat Won WSI-Client 2012 BaseClient WonBaseClient DemoStar Lost WSI-Client 2012 Spartiat LostBaseClient WSI-Client 2012 Lost WSI-Client 2012 DemoStar TieBaseClient KevinCar Lost WSI-Client 2012 KevinCar Lost SloppyJalopyBaseClient SEALbot Lost WSI-Client 2012 SEALbot Tie SloppyJalopy BaseClient LostBaseClient SloppyJalopy Won WSI-Client 2012 SloppyJalopy Won SloppyJalopy Spartiat WonBaseClient JustDetermined Won WSI-Client 2012 JustDetermined Won SloppyJalopy DemoStar LostMatches Won: 3 Matches Won: 3 SloppyJalopy WSI-Client 2012 LostSpartiat KevinCar SloppyJalopy KevinCar LostSpartiat BaseClient Lost KevinCar BaseClient Won SloppyJalopy SEALbot TieSpartiat DemoStar Tie KevinCar Spartiat Won SloppyJalopy JustDetermined LostSpartiat WSI-Client 2012 Won KevinCar DemoStar Won Matches Won: 1Spartiat KevinCar Lost KevinCar WSI-Client 2012 WonSpartiat SEALbot Lost KevinCar SEALbot Tie JustDeterminedSpartiat SloppyJalopy Lost KevinCar SloppyJalopy Won JustDetermined BaseClient LostSpartiat JustDetermined Won KevinCar JustDetermined Won Lost JustDetermined SpartiatMatches Won: 2 Matches Won: 6 JustDetermined DemoStar Tie JustDetermined WSI-Client 2012 LostDemoStar SEALbot Won JustDetermined KevinCar LostDemoStar BaseClient SEALbot BaseClient WonDemoStar Spartiat Tie SEALbot Spartiat Won JustDetermined SEALbot LostDemoStar WSI-Client 2012 Tie SEALbot DemoStar Lost JustDetermined SloppyJalopy WonDemoStar KevinCar Lost SEALbot WSI-Client 2012 Tie Matches Won: 1DemoStar SEALbot Won SEALbot KevinCar TieDemoStar SloppyJalopy Won SEALbot SloppyJalopy TieDemoStar JustDetermined Tie SEALbot JustDetermined WonMatches Won: 3 Matches Won 3 26 © 2012 University of Tuebingen
  • Preliminary 1 vs. 1 MatchesRanking in Preliminaries…. 1. KevinCar 6 2. BaseClient 3 DemoStar 3 SEALbot 3 WSI-Client 2012 3 6. Spartiat 2 7. SloppyJalopy 1 JustDetermined 127 © 2012 University of Tuebingen
  • Final Show-Down• All eights controllers run against each other• 10 runs.• If one controllers is our (full damage) all other damages are reset to zero!• Results: - Close but there is a clear winner…28 © 2012 University of Tuebingen
  • 29 © 2012 University of Tuebingen
  • And the Winner is…Run 1 Points Run 4 Run 71st WSI-Client 2012 7 1st SEALbot 7 1st KevinCar 72nd SloppyJalopy 6 2nd SloppyJalopy 6 2nd BaseClient 63rd SEALbot 5 3rd KevinCar 5 3rd SEALbot 54th KevinCar 4 4th DemoStar 4 4th WSI-Client 2012 45th BaseClient 3 5th WSI-Client 2012 3 5th SloppyJalopy 36th DemoStar 2 6th BaseClient 2 6th DemoStar 27th JustDetermined 1 7th JustDetermined 1 7th Spartiat 18th Spartiat 0 8th Spartiat 0 8th JustDetermined 0Run 2 Run 5 Run 81st SEALbot 7 1st KevinCar 7 1st SEALbot 72nd BaseClient 6 2nd BaseClient 6 2nd SloppyJalopy 63rd WSI-Client 2012 5 3rd SEALbot 5 3rd BaseClient 54th KevinCar 4 4th SloppyJalopy 4 4th KevinCar 45th SloppyJalopy 3 5th WSI-Client 2012 3 5th DemoStar 36th JustDetermined 2 6th DemoStar 2 6th JustDetermined 27th Spartiat 1 7th Spartiat 1 7th Spartiat 18th DemoStar 0 8th JustDetermined 0 8th WSI-Client 2012 0Run 3 Run 6 Points Run 91st JustDetermined 7 1st BaseClient 7 1st SEALbot 72nd KevinCar 6 2nd SloppyJalopy 6 2nd SloppyJalopy 63rd SloppyJalopy 5 3rd DemoStar 5 3rd WSI-Client 2012 54th WSI-Client 2012 4 4th JustDetermined 4 4th DemoStar 45th DemoStar 3 5th SEALbot 3 5th Spartiat 36th BaseClient 2 6th Spartiat 2 6th KevinCar 27th Spartiat 1 7th WSI-Client 2012 1 7th BaseClient 18th SEALbot 0 8th KevinCar 0 8th JustDetermined 030 © 2012 University of Tuebingen
  • And the Winner is…Run 1 Points Run 4 Run 71st WSI-Client 2012 7 1st SEALbot 7 1st KevinCar 72nd SloppyJalopy 6 2nd SloppyJalopy 6 2nd BaseClient 63rd SEALbot 5 3rd KevinCar 5 3rd SEALbot 54th KevinCar 4 4th DemoStar 4 4th WSI-Client 2012 45th BaseClient 3 5th WSI-Client 2012 3 5th SloppyJalopy 36th DemoStar 2 6th BaseClient 2 6th DemoStar 27th JustDetermined 1 7th JustDetermined 1 7th Spartiat 18th Spartiat 0 8th Spartiat 0 8th JustDetermined 0 Run 10Run 2 Run 5 Run 8 1st SEALbot 71st SEALbot 7 1st KevinCar 7 1st SEALbot 7 2nd JustDetermined 6 2nd SloppyJalopy 62nd BaseClient 6 2nd BaseClient 63rd WSI-Client 2012 5 3rd3rd SloppyJalopy SEALbot 5 5 3rd BaseClient 54th KevinCar 4 4th4th SloppyJalopy KevinCar 4 4 4th KevinCar 45th SloppyJalopy 3 5th5th WSI-Client 2012 WSI-Client 2012 3 3 5th DemoStar 36th JustDetermined 2 6th DemoStar 2 6th JustDetermined 2 6th BaseClient 27th Spartiat 1 7th Spartiat 1 7th Spartiat 1 7th Spartiat 18th DemoStar 0 8th JustDetermined 0 8th WSI-Client 2012 0 8th DemoStar 0Run 3 Run 6 Points Run 91st JustDetermined 7 1st BaseClient 7 1st SEALbot 72nd KevinCar 6 2nd SloppyJalopy 6 2nd SloppyJalopy 63rd SloppyJalopy 5 3rd DemoStar 5 3rd WSI-Client 2012 54th WSI-Client 2012 4 4th JustDetermined 4 4th DemoStar 45th DemoStar 3 5th SEALbot 3 5th Spartiat 36th BaseClient 2 6th Spartiat 2 6th KevinCar 27th Spartiat 1 7th WSI-Client 2012 1 7th BaseClient 18th SEALbot 0 8th KevinCar 0 8th JustDetermined 031 © 2012 University of Tuebingen
  • And the Winner is.... SEALbot Anderson Rocha Tavares Anderson Rocha Tavares & Gabriel de Oliveira Ramos & Renato de Pontes Pereira & Sérgio Montazzolli Silva & Ana L. C. BazzanUniversidade Federal do Rio Grande do Sul (UFRGS), Porto Alegre UFRGS Brasil
  • Final Scores and Rankings Rank Name # of Wins Points 1 SEALbot 5 53 3 KevinCar 2 43 4 BaseClient 1 40 5 WSI-Client 2012 1 35 6 JustDetermined 1 32 2 SloppyJalopy 0 50 7 DemoStar 0 25 8 Spartiat 0 1133 © 2012 University of Tuebingen
  • Institute of Computer Science Chair of Cognitive ModelingThank you for the attention!