Difference between revisions of "ScalarX"

From Robowiki
Jump to navigation Jump to search
m (open source bot-development framework, with machine learning)
m (Debug System, record battles even in roborunner & roborumble, with full replay capacity.)
 
Line 10: Line 10:
 
* Training
 
* Training
 
** A complete pipeline for running battles, logging data, offline training and testing.
 
** A complete pipeline for running battles, logging data, offline training and testing.
 +
* Debugging
 +
** A complete system to record battles (even in [[RoboRunner]] and [[RoboRumble]]), and replay in robocode like a video with backward, forward, slo-mo, etc.
 
* To be listed...
 
* To be listed...

Latest revision as of 06:21, 16 July 2021

ScalarX is an open source bot-development framework, aims at easing the gap between robocode and modern machine learning techniques. Currently building a competitive robot requires months of coding, and yet many details remain to be tested. If there is a reference implementation which can be tested exclusively by the community, and let any strategy & machine learning model plug in easily, work of exploring & applying modern machine learning techniques can be greatly simplified.

Current development is in prototype phase, let me know if you are interested.

Components

ScalarX will include:

  • A utility library for implementing things like KNN, Precise Prediction, Bullet Shadow, etc.
  • Serving
    • A component system to enable easy plug-in, e.g. shield, multiple movement system, virtual guns, etc.
  • Training
    • A complete pipeline for running battles, logging data, offline training and testing.
  • Debugging
    • A complete system to record battles (even in RoboRunner and RoboRumble), and replay in robocode like a video with backward, forward, slo-mo, etc.
  • To be listed...