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

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)

• 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