DeepMind, a UK based Artificial Intelligence (AI) company and research laboratory owned by Alphabet Inc, just open-sourced Lab2D, an AI learning environment written in C++ and Lua for the creation of grid worlds.
The main goals of the system are ease of use and performance: The environments are “grid worlds”, which are defined with a combination of simple text-based maps for the layout of the world, and Lua code for its behavior. The system is tailored to multi-agent reinforcement learning.
Two-dimensional environments are inherently easier to understand than three-dimensional ones, at very little, if any, loss of expressiveness. Even a game as simple as Pong, which essentially consists of three moving rectangles on a black background, can capture something fundamental about the real game of table tennis. This abstraction makes it easier to capture the essence of real-world problems and concepts that need to be solved.
DeepMind Lab2D provides several flexible mechanisms for exposing internal environment information: The simplest form is through observations, which allow the researcher to add specific information from the environment to the observations that are produced at each time step. The second way is through events, which, similar to observations, can be raised from within the Lua script.
The researchers argue in their paper, “Rich complexity along numerous dimensions can be studied in 2D just as readily as in 3D, if not more so… In addition, 2D worlds are significantly less resource-intensive to run, and typically do not require any specialized hardware (like GPUs) to attain reasonable performance.”