Enhance RL with differentiable physics

This is a submodule of our differentiable articulated body simulator. Please refer to the main repo for more details.

Acknowledgments

Most of the code in this repo is derived from MBPO (When to Trust Your Model: Model-Based Policy Optimization).