Skip to content
/ Kokomo Public

Kokomo is a competitor robot built for Robocode that uses regression tree based machine learning (called "Dynamic Segmentation" on the Robocode wiki) for its targeting strategy

Notifications You must be signed in to change notification settings

0x65-e/Kokomo

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

7 Commits
 
 
 
 
 
 

Repository files navigation

Kokomo

Kokomo is a competitor robot built for Robocode that uses regression tree based machine learning (called Dynamic Segmentation by some in the Robocode community) to learn its opponent's movement strategy.

Like most top competitor robots, Kokomo collects GuessFactors using Waves - both its own and its opponents. Movement uses traditional Wave Surfing (a slight improvement over BasicGFSurfer's reference Wave Surfing implementation). Targeting uses a custom incremental regression tree.

Future Plans

  • Improve the performance of the regression tree and solve several memory issues that occur when matches go on too long
  • Implement serialization, like described in the Wiki Targeting page, to improve early round performance against top bots
  • Expand regression tree into a forest, starting with an implementation of a k-d tree and a k-nn targeting algorithm
  • Investigate using machine learning for movement as well

About

Kokomo is a competitor robot built for Robocode that uses regression tree based machine learning (called "Dynamic Segmentation" on the Robocode wiki) for its targeting strategy

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages