Fork of the Arcade Learning Environment (ALE)

For the main (upstream) project see: https://github.com/mgbellemare/Arcade-Learning-Environment

The changes here are intended to:

• Provide a serializable state that represents unique game states
• Apply fixes that haven't yet made it into upstream
• Make it easier to "vendor" the ALE source under another C++ project and compile it directly (vs building ALE as a library)

Specific changes from upstream:

• Paddle positions are exposed in the API (needed for complete observable state)
• Applied Jesse Farebrother's patch to fix nondeterminism issues
  • Upstream pull request: Farama-Foundation#401
• Only the base C++ interface; no Python bindings
• No CMake or Docker