stanfordnmbl / osim-rl

Reinforcement learning environments with musculoskeletal models
http://osim-rl.stanford.edu/
MIT License
882 stars 249 forks source link

Create a new environment #169

Closed epris closed 5 years ago

epris commented 5 years ago

Dear all,

We are evaluating to use OpenSim and Osim-rl for our research (In the past we used another framework).

If I understand well, at the moment osim-rl is equipped with 3 environments: the arm, the NIPS 2017 and the NIPS 2018.

I want to know if It is possible for an user to add a new environment starting from one of the models created for OpenSim and available on simtk. (I mean on my local machine. Is the use of crowdAI useful only for the competition?)

Why is Arm2dEnv considered a Toy model? For the number of muscles or for the model of muscles?

Best regards, Antonio

kidzik commented 5 years ago

That's true, crowdAI part is only for the competition and you can use our library for other projects. In order to implement a new model you basically need to:

We call the Arm2dEnv a toy model because the range of motion is very limited. Also, it was designed for throwing something rather than for grasping.