Repos

MuJoCo - https://github.com/osudrl/cassie-mujoco-sim

Gazebo - https://github.com/agilityrobotics/cassie-gazebo-sim

About the simulators

We offer two dynamic simulations of Cassie that you can use to evaluate the physical capabilities of the robot and try out your controls ideas. The sims are continuously validated against the real robot and are what we use to develop our own controllers. Our customers have had excellent results developing against the sims prior to their robots shipping, and experience little to no downtime switching to the actual robot hardware.

In addition to trying controls ideas in simulation first, the simulators provide a means to try out methods of control that Agility does not support in-house, such as deep learning. Because the sims are a close match to the actual physics of the robot, operations that would require considerable time (and risk of damage) in the real world can be performed in simulation first and later deployed to real hardware. In the video below, our collaborators at the University of British Columbia and Oregon State University show their reinforcement learning results as developed in simulation.