google-deepmind/dm_control is Google DeepMind's software stack for physics-based simulation and reinforcement learning environments, built on MuJoCo. The latest release is 1.0.49, and the changes carried in this version are not recorded in the collected data.
It provides a simulation-based platform for running and observing robotics and physics control tasks. The core components are dm_control.mujoco, which offers Python bindings, dm_control.suite, a collection of reinforcement learning environments, and dm_control.viewer, an interactive viewer. It also includes dm_control.mjcf, dm_control.composer, dm_control.locomotion and dm_control.locomotion.soccer for building complex control tasks, and the README explains installation and rendering setup. The license is Apache-2.0 and the primary language is Python. It has 4,715 GitHub stars, and the repository was created on December 29, 2017.
However, the public information stops at the environment setup, the license and the repository numbers. There is no changelog for 1.0.49, and no performance figures or benchmarks are presented. In the rating out of 10, the momentum item remains low at 1.7, while the freshness, technique and quality items are empty. The environment setup, the viewer and the control task components are worth referencing, but what this version changed and how it performs in practice are hard to judge until the release records are checked directly.