Why isn't there a published TF frame with tray frame?

[Ross Wightman]

Up until this point I was assuming agv1/2_load_point_frame was a frame I could use for parts placement given they are the only published AGV frames and referenced in the materials location interface. As I noticed this was putting parts out of reach of the arm I looked into it more and determined it wasn't usable for that purpose. The actual agv1::kit_tray::kit_tray::tray co-ordinates are different and what the parts orders reference.

Why isn't there an easily accessible TF frame for the tray and parts orders co-ordinates? I'm not extremely familiar with ROS and can't find a clean way of accessing the co-ordinates of the agv1::kit_tray::kit_tray::tray via Python moduels so I've resorted to manually offsetting the load point TF by the tray offset using a constant...

[Deanna Hood]

Hi @rwightman-hwl, in general, the only TF frames that are published by the ARIAC simulation without any sensors are reference points to areas in the workcell that do not change over time. This includes the location of storage units such as bins and the loading point for AGVs.

To determine the location of dynamic objects, such as movable parts, you are expected to use sensors. Something that will become clearer in the coming qualification tasks is that the AGVs are also dynamic -- they deliver trays and return with empty trays (but not in qual1). As such, the kit trays on AGVs fall under the category of movable parts. This is why the location of kit trays is not published by default: it is supposed to be determined by the use of sensors. You can, for example, mount a logical camera above the AGV load point which will then publish the TF frame of the kit tray (for qual1 it is 'kit_tray_frame' but from qual2 it will be a unique frame name).

Thank you for asking this question -- you are not the first to raise this issue (https://discourse.ros.org/t/part-placement/1361/3), so we will be sure to clarify this in the competition documentation.

[Ross Wightman]

Thanks for the response. I expected the tray would be dynamic in future tasks but didn't realize that deploying sensors and paying that price was expected to obtain that information vs the AGV being responsible for reporting its current configuration/state and thus publishing the tray TF as say the arm link frames are published.

[Deanna Hood]

I can understand that you would expect this information to be made available without the need for sensors. In future tasks you will most likely find that sensors above the trays are necessary anyway, for quality control. Should anything change with regards to accessing information about kit tray poses, we will let you and other participants know, but for now I will mark this question as resolved.