Third-party environments

Here, we feature a selection of third-party libraries that build upon or complement ViZDoom, offering diverse environments and tools for reinforcement learning research and development.

Please note that the page contains environments that are not maintained by the ViZDoom Team or Farama Foundation.

If you have a library that you would like to see featured here, please open a pull request or an issue on the GitHub repository

LevDoom

LevDoom is a benchmark for generalization in pixel-based deep reinforcement learning, offering environments with difficulty levels based on visual and gameplay modifications. It consists of 4 scenarios, each with 5 difficulty levels, that modify different aspects of the base environments, such as textures, obstacles, types, sizes, and rendering of different in-game entities, etc.

LevDoom provides environments using Gymnasium API and is available through PyPi. For more details, please refer to the CoG2022 paper and GitHub repository.

COOM

COOM is a Continual Learning benchmark for embodied pixel-based RL, consisting of task sequences in visually distinct 3D environments with diverse objectives and egocentric perception. COOM is designed for task-incremental learning, in which task boundaries are clearly defined. It contains 8 scenarios, every with at least 2 difficulty levels that are combined into sequences of tasks for Continual Learning. The sequence length varies between 4, 8, and 16. COOM provides two types of task sequences:

  • Cross-domain sequences compose modified versions of the same scenario (e.g., changing textures, enemy types, view height, and adding obstacles) while keeping the objective consistent.

  • Cross-objective sequences contrast with Cross-Domain by changing both the visual characteristics and the objective for each task, requiring a more general policy from the agent for adequate performance.

COOM provides environments using Gymnasium API and is available through PyPi. For more details, please refer to the GitHub repository.