mjlab
mjlab is a robot reinforcement learning training framework built on MuJoCo, used for training and evaluating control policies for humanoid robots.
Timeline (last 90 days)
Developers are building a whole-body recovery policy for the Asimov 1 humanoid robot, with the goal of recovering from a random supine, prone, or side-lying posture (with joint perturbations) to an upright and stable stance.
The code and training setup have been made public on the feature/mjlab-standup branch of the GitHub repository.
microduck_rl is a reinforcement learning training environment built specifically for the Microduck robot, constructed on top of mjlab and developed primarily in Python.
The project's policy was trained from scratch using mjlab (a MuJoCo wrapper tool) and converged in approximately 4 hours on a single RTX 3080.
The policy was trained from scratch using mjlab (a wrapper for MuJoCo) and converged in approximately 4 hours on a single RTX 3080.
All Facts (5)
Developers are building a whole-body recovery policy for the Asimov 1 humanoid robot, with the goal of recovering from a random supine, prone, or side-lying posture (with joint perturbations) to an upright and stable stance.
50%Unverifiedmicroduck_rl is a reinforcement learning training environment built specifically for the Microduck robot, constructed on top of mjlab and developed primarily in Python.
50%UnverifiedThe project's policy was trained from scratch using mjlab (a MuJoCo wrapper tool) and converged in approximately 4 hours on a single RTX 3080.
50%UnverifiedThe policy was trained from scratch using mjlab (a wrapper for MuJoCo) and converged in approximately 4 hours on a single RTX 3080.
50%UnverifiedThe code and training setup have been made public on the feature/mjlab-standup branch of the GitHub repository.
50%