Publishing odometry with valid covariances

[Imaniac230]

Hello,

the odometry message currently published by the driver doesn't include any covariance estimates. This means that when using other downstream fusion algorithms (like robot_localization) they have to be estimated somehow experimentally (in addition to the issue discussed here: #29).

The SDK docs for the proto API, however, do specify a pose covariance message here but I didn't find any use of it in the ros2 driver code. Does the robot output any valid estimates in this message a would it be possible to acquire the data and pass it into the ROS odometry message?

Especially when using the vision-based odometry from robot, I think it would be a more precise to have it coming directly from the internal algorithm than any second-hand experimental estimates.

[khughes-bdai]

Looking at the sdk docs, it looks like the only places this covariance message is exposed is through graph nav waypoint detection and fiducial detection. I unfortunately don't think it's possible to get the covariance of the robot's odometry pose itself through the sdk currently. This could be a good issue to raise with Boston Dynamics customer support as I agree it would be best to have the covariance directly from the robot.

[Imaniac230]

Thanks for the response. Just to follow up, I opened a support ticket with them, but got a negative response. So, sadly, I don't think that this is going to be a thing anytime soon.