Nikita Rudin

Thanks for reporting. It should be `HeightFieldParams`. Feel free to do a pull request if this fixes the issue.

terran_kwargs needs to be a dict of keyword arguments for that particular terrain. For example: `env_cfg.terrain.terrain_kwargs = { 'type': 'pyramid_stairs_terrain', 'step_width': 0.31, 'step_height': -0.1, 'platform_size': 3.}`

This might be related to your hardware (memory limitation for example). I have verified that latest pytorch + cuda work on a wide range 20 and 30 series GPUs

Hi, The privileged observations feature is implemented but were not using it in the paper. These privileged observations are not used in a teacher-student distillation. Instead, they are used in...

the equivalent in the tensor api is `rigid_body_state_tensor = self.gym.acquire_rigid_body_state_tensor(self.sim)` for reading the values and `self.gym.set_actor_root_state_tensor_indexed` for setting root states. You should be able to use those to set positions...

You should be able to add shifting platforms as boxes, but our current height scanner won't work with that since we get the heights from the static terrain.

Can you check that you are using Isaac Gym preview 3 and not preview 2? Our URDFs have an additional tag that is only supported by preview 3.

You can do a grid search by running the training in a for loop and changing the configs for each iteration. The only caveat is that you need to run...