Quick Start

After you install MyoAssist (see Getting Started), run one of these minimal scripts to confirm your setup builds and steps a composed environment. Neither one trains a policy, they only create an environment and run it.

Minimal RL environment

This creates a Gym-wrapped MuJoCo environment and steps it with random actions for about 150 frames (5 seconds).

python rl_train/run_sim_minimal.py

On macOS, use mjpython so the MuJoCo viewer works:

mjpython rl_train/run_sim_minimal.py

Minimal reflex controller (CO)

This builds a composed reflex environment, runs a random controller, and prints how long the model stays upright. The walking duration will likely change each run, because the control parameters are random.

python ctrl_optim/run_ctrl_minimal.py

Before you train

Set MYOASSIST_CACHE_DIR first. The model is composed at run time, so a training run without the cache is much slower. See Step 5 of the installation.

Next steps

• Defining an Environment: choose the MSK model, device, and terrain to simulate.
• Examples: ready-to-run environment specs.
• Reinforcement Learning: train a policy.
• Controller Optimization: optimize a reflex controller.