Add six new environments: BlocksRanking, StackCups, StackBlocksTwo, MatchObjectContainer, and PlaceObjectDrawer #70
Conversation
| "uid": "red_block_mat" | ||
| } | ||
| }, | ||
| "set_block_2_color": { |
There was a problem hiding this comment.
If we don't randomize the material params, we can use startup mode for this functor
| "sim_update": true | ||
| } | ||
| }, | ||
| "random_material": { |
There was a problem hiding this comment.
We should keep basic skill task env simple. Just remove the following randomization functors
| "entity_cfg": {"uid": "block_3"} | ||
| } | ||
| }, | ||
| "cam_high_semantic_mask_l": { |
There was a problem hiding this comment.
Same, we don't need these obs
| } | ||
| } | ||
| }, | ||
| "dataset": { |
There was a problem hiding this comment.
Refer to the updated dataset recorder example in configs/pour_water/gym_config.json
There was a problem hiding this comment.
the key "dataset" can be removed from gym_config.json if there's no need to save dataset
| }, | ||
| "sensor": [ | ||
| { | ||
| "sensor_type": "StereoCamera", |
There was a problem hiding this comment.
We may follow the common setup of cobotmagic robot (3 camera with the similar params)
| } | ||
| ] | ||
| } | ||
|
|
There was a problem hiding this comment.
Change the gym_config.json into cobot_magic_3cam.json to explicitly indicate that the robot and sensor setup. Also we may have different embodiment setup for one task
| @@ -0,0 +1,182 @@ | |||
| # ---------------------------------------------------------------------------- | |||
There was a problem hiding this comment.
Maybe geometry is better for the file name
| ) | ||
|
|
||
|
|
||
| def randomize_rigid_object_body_scale( |
There was a problem hiding this comment.
Could we remove these two functors if they will not be used anymore
| env: EmbodiedEnv, | ||
| env_ids: Union[torch.Tensor, None], | ||
| entity_cfg: SceneEntityCfg, | ||
| base_color: list[float] | None = None, |
There was a problem hiding this comment.
Could we just use VisualMaterialCfg? It can also created from dict
|
|
||
|
|
||
| @register_env("BlocksRankingRGB-v1", max_episode_steps=600) | ||
| class BlocksRankingRGBEnv(EmbodiedEnv): |
There was a problem hiding this comment.
For these six task, we may create a simple expert trajectory for demonstration
Description
This PR adds six new environments to
EmbodiChain.1. BlocksRankingRGB-v1 (
BlocksRankingRGBEnv)2. BlocksRankingSize-v1 (
BlocksRankingSizeEnv)3. StackCups-v1 (
StackCupsEnv)4. StackBlocksTwo-v1 (
StackBlocksTwoEnv)5. MatchObjectContainer-v1 (
MatchObjectContainerEnv)container_cube.container_sphere.6. PlaceObjectDrawer-v1 (
PlaceObjectDrawerEnv)inner_boxarea.Type of change
Screenshots
Checklist
black .command to format the code base.