Advanced features

There are two main components of the osim-rl:

  • OsimModel class responsible for low level communication with OpenSim simulator.
  • OsimEnv class responsible for management of the user interface for running reinforcement learning algorithms.
Unreleased

This page is under construction.

Methods of OsimModel

list_elements()

List names of model elements including: joints, bodies, muscles, forces, and markers.

get_body(name), get_joint(name), get_muscle(name), get_marker(name), get_force(name)

Get an element by name. To better understand what can you do with these entities, please refer to OpenSim documentation. You can find list of all function under the given class, for example, for muscles please refer to this file.

actuate(action)

Set muscle excitations.

set_activations(action)

Set muscle activations directly in the state.

get_activations(action)

Get muscle activations in the current state.

get_state_desc(action)

Get a dictionary describing the state of joints, bodies, muscles, forces, and markers.

get_state()

Get the current state of the environment.

set_state(state)

Set the state to state.

integrate()

Run one step of the simulation.

Methods of OsimEnv

You can access the OsimModel associated with OsimEnv you can access env.model.