Rerun fixes

[leshy]

urdf

# URDF path for Go2 robot
_GO2_URDF = Path(__file__).parent.parent / "go2" / "go2.urdf"


        # Load robot URDF
        if _GO2_URDF.exists():
            rr.log_file_from_path(
                str(_GO2_URDF),
                entity_path_prefix="world/robot",
                static=True,
            )
            logger.info(f"Loaded URDF from {_GO2_URDF}")

this was done because foxglove is dumb and won't let me draw a box in 3d
with rerun you can, so no need for this fake box urdf, just draw the box.

TF module

idk why we have in blueprints:

    tf_rerun(),  # Auto-visualize all TF transforms in Rerun

but then we build/vis all the transforms manually

        # Log camera transform (compose base_link -> camera_link -> camera_optical)
        # transforms[1] is camera_link, transforms[2] is camera_optical
        cam_tf = transforms[1] + transforms[2]  # compose transforms
        rr.log(
            "world/robot/camera",
            rr.Transform3D(
                translation=[cam_tf.translation.x, cam_tf.translation.y, cam_tf.translation.z],
                rotation=rr.Quaternion(
                    xyzw=[
                        cam_tf.rotation.x,
                        cam_tf.rotation.y,
                        cam_tf.rotation.z,
                        cam_tf.rotation.w,
                    ]
                ),
            ),
        )


        # Log to Rerun: robot pose (relative to parent entity "world")
        rr.log(
            "world/robot",
            rr.Transform3D(
                translation=[msg.x, msg.y, msg.z],
                rotation=rr.Quaternion(
                    xyzw=[
                        msg.orientation.x,
                        msg.orientation.y,
                        msg.orientation.z,
                        msg.orientation.w,
                    ]
                ),
            ),
        )
        # Log axes as a child entity for visualization
        rr.log("world/robot/axes", rr.TransformAxes3D(0.5))  # type: ignore[attr-defined]

we can't do any of this stuff in go2 connection.
at most (if not using a rerun tf module) it should be this:

self.tf.publish(*transforms)

for transform in transforms:
    rr.log(.... transform.to_rerun())

actual rerun transform module should be similar to above also..

never build objects manually that are equivalent to dimos objects - create to_rerun() methods

camera image distortion

camera image needs to be logged under a projection so that it takes camerainfo distortion into account

tf module

            self._lcm = LCM()
            self._lcm.start()
            topic = Topic("/tf", TFMessage)

too much vibing, you need to understand things, and ask if you don't understand

we have self.tf module for transforms, don't touch lcm directly

[linear]

DIM-329