Support TwistWithCovarianceStamped

[JonasHablitzel]

Hi,

i wanted to ask if there is any plan to publish TwistWithCovarianceStamped instead of TwistStamped for the ins/velocity Topic, similar to the ROS1 driver?

[ljones-oxts]

Hi Jonas

There are no plans at the moment to publish TwistWithCovarianceStamped - part of the issue is that the angular rate covariances are not actually stored in our NCOM format.

I can raise the issue internally if this is something that is causing you issues.

Thanks

Llyr

[JonasHablitzel]

Hi Llyr,

thank for the quick response, we plan to use the OXTS together with the Autoware stack. As Autoware expects TwistWithCovarianceStamped an adaptation would be neccesary. Do you think an implementation of the usecase on the ROS-Driver Side would be also possible? So we could avoid an adaptation of the NCOM format?

[marioney]

We're facing this same issue. We worked around by changing the topic type and setting a fixed value for the covariance, but of course the correct value for the is still missing.

@JonasHablitzel did you find a different solution for this?
@ljones-oxts Has there been any progress for this.

[ljones-oxts]

Hi @marioney and @JonasHablitzel

My apologies for neglecting this thread. Having discussed with my colleagues, we are unable to provide the actual covariance measurements for twist (due to internal IP concerns). However, would your tool be able to accept a TwistWithCovarianceStamped message even if the covariance entries are set to zeroes (or some other dummy value)?

Thanks

Llyr

[marioney]

Thanks for your response @ljones-oxts

I think it should work with dummy covariance value, zeroes may not be the best option, I'd prefer to use a fixed "average" covariance value, or maybe something that can be configured by a parameter.