Question About Station-Keeping Setup

[itserichuynh]

Hi Evan,

I recently read your paper and was particularly interested in the station-keeping and waypoint navigation setup used in your experiments. I was wondering—did you use an ArUco tag for station-keeping, or was the Qualisys motion capture system providing localization?

Our team is planning a similar setup, so any guidance on this part would be greatly appreciated. From a quick look at the blue_localization package, it seems there’s an option to select between camera or Qualisys-based localization, and I’m curious which you used in the paper’s experiments. Also, did you use the whole-body controller for station-keeping in that setup?

Thanks

[evan-palmer]

Hey Eric,

For those experiments we used a motion capture system to provide localization. Our access to that facility is limited, so these days we use fiducials for pose estimation (see e.g., barlus_underwater_camera). We fuse those estimates with an IMU using an EKF. For our field deployments, we use the Water Linked USBL and DVL. I'll note that the blue_localization package is in the process of being deprecated. The feat_deprecate_localization has the latest updates, including demos for the full UVMS. That branch should be working, it is currently held up by structural documentation changes that need to be made but I haven't had time for yet.

Regarding the station-keeping experiments, yes we used the whole-body controller. All of the control algorithms from that paper are available in auv_controllers.

[itserichuynh]

Hi Evan,

Thanks so much for the quick response!

I took a closer look at the whole-body controller setup for UVMS, and I see that it requires both the global pose of the robot and the target. I was wondering if it would work at all if we only had the local position of a target relative to the robot?

Our current idea is to station-keep the BlueROV with respect to an AprilTag, so I’m curious if that would fit within the existing control framework.

Thanks again for your help,

Eric

[evan-palmer]

The full configuration of the UVMS, including the pose of the vehicle is needed for the IK solver (see here). Regarding the reference position being in a local frame, that should work, depending on your setup. For example, if you are estimating the pose of the vehicle with respect to an AprilTag using a camera positioned on the vehicle and subsequently define your reference pose with respect to that same camera, you can just transform your reference pose into the global (marker) frame using the current vehicle pose and send that to the controller.