Christoph Wiesmeier

This page is read only. You can view the source, but not change it. Ask your administrator if you think this is wrong.

====== SuperHexagon AI ======

  * Video of the Game: [[https://www.youtube.com/watch?v=5mDjFdetU28]]
  * Video of an algorithm based Player [[https://www.youtube.com/watch?v=EUxRYix2aaM]]

===== Goal =====
 
Let the computer play the game.

  * Vision-Based, the only input it the screen image. The game is controlled via keyboard emulation.
  * Use a deep learning framework to train an artificial player.
  * Compare different approaches. In terms of leaning rate and success.
    * Supervised Learning
    * Reinforcement Learning
    * Algorithmic Approach
  * Compare the influence of preprocessing steps e.g.
    * Raw image
    * Polar Transformation
    * Rotated (player has fixed position)

===== Needed Steps =====

  * Create/Extend the current test framework.
    * Automatically start the game
    * Evaluate the results
    * Compare different Implementations
    * Robustness
  * Select a suitable machine learning framework
  * Collect data from human player and algorithm-based implementation
  * Train a learning model
  * Improve on debugging and Analytics tools.
  * Iterate on the design
  * Collect and analyze the Result