[Question] Differentiate objects when using MultiAssetSpawnerCfg

[luju97]

Hi,

I am usingMultiAssetSpawnerCfg to spawn different objects in each environment.

I'm trying to train my robotic arm to pick up different types of objects, starting from the FrankaCubeLiftEnvCfg example.

How can I differentiate these objects during training? (I want to add this info to the observation space)
Is there a way to access information such as the usd path? I am unable to differentiate them since they are all under the same prim path.

Thank you in advance!

[RandomOakForest]

Thanks for posting this. Check out this tutorial on spawning multiple assets and let us know if you need further help.

[luju97]

@RandomOakForest I am using MultiAssetSpawnerCfg, like in the example:

object: RigidObjectCfg = RigidObjectCfg(
    prim_path="/World/envs/env_.*/Object",
    spawn=sim_utils.MultiAssetSpawnerCfg(
        assets_cfg=[
            sim_utils.ConeCfg(
                radius=0.3,
                height=0.6,
                visual_material=sim_utils.PreviewSurfaceCfg(diffuse_color=(0.0, 1.0, 0.0), metallic=0.2),
            ),
            sim_utils.CuboidCfg(
                size=(0.3, 0.3, 0.3),
                visual_material=sim_utils.PreviewSurfaceCfg(diffuse_color=(1.0, 0.0, 0.0), metallic=0.2),
            ),
            sim_utils.SphereCfg(
                radius=0.3,
                visual_material=sim_utils.PreviewSurfaceCfg(diffuse_color=(0.0, 0.0, 1.0), metallic=0.2),
            ),
        ],
        random_choice=True,
        rigid_props=sim_utils.RigidBodyPropertiesCfg(
            solver_position_iteration_count=4, solver_velocity_iteration_count=0
        ),
        mass_props=sim_utils.MassPropertiesCfg(mass=1.0),
        collision_props=sim_utils.CollisionPropertiesCfg(),
    ),
    init_state=RigidObjectCfg.InitialStateCfg(pos=(0.0, 0.0, 2.0)),
)

I would like to know if there is a way to know whether the spawned object is a cone, a cube or a sphere (without using vision systems), just by getting this info from the simulator.

Thank you

[RandomOakForest]

Is the semantic_tags attribute something you could use?

[luju97]

@RandomOakForest If I use semantic_tags how can I access this info to add it to my observation space? E.g. something like a tensor of the tags for each environment.

[RandomOakForest]

Following up on this, let us know how that went. We have a show and tell section where you can post findings like this. I'll move this issue to our Discussions section for follow up in the meantime.

[RandomOakForest]

Semantic tags may be a long road for this, as this may require a sensor and use of Replicator to get the annotations you need. Alternatively, you can use a contact sensor to get the object info you are making contact with.