Standard Message for Sensor Accuracy

asked 2019-10-21 14:42:24 -0500

SChung gravatar image

We are using an NDI measurement system that returns an "Indicator Value" for how much error is estimated on a reading. This is a single value that corresponds with the goodness of fit for 6-DOF sensors.

Indicator Value: X . XXXX

An estimate of how well the Aurora System calculated the transformation. Values range from 0 to 9.9. A higher value indicates a higher error.

Is there a message type that is typically used for these values? Nothing jumped out when looking at the geometry_msgs and sensor_msgs. My default thought would just be a generic Float64

edit retag flag offensive close merge delete


measurement system that returns an "Indicator Value" for how much error is estimated on a reading.

Would this be the (co)variance associated with the measurement?

If so: there are other message types which have a field that encodes for that. sensor_msgs/Temperature is one of them. It would indeed "just" be a float64, but given the name and the context that doesn't seem so bad.

gvdhoorn gravatar image gvdhoorn  ( 2019-10-21 14:47:50 -0500 )edit

The documentation is a bit vague for this system. I imagine some combined metric of variance of the sensor's orientation and translation for simplicity. I would feel okay about the float message type, but just needed a second opinion!

SChung gravatar image SChung  ( 2019-10-21 15:25:42 -0500 )edit

I'm be surprised if they didn't also expose the covariances used to create this metric

stevemacenski gravatar image stevemacenski  ( 2019-10-21 15:51:40 -0500 )edit

I searched through the API doc and didn't see any additional tracking information to request. I'll see if an email turns up anything I may have missed.

SChung gravatar image SChung  ( 2019-10-21 16:08:53 -0500 )edit

Could it perhaps some sort of inverted confidence indication? So 0.0 is "full confidence", 9.9 is "no confidence". Almost a percentage divided by 10.

gvdhoorn gravatar image gvdhoorn  ( 2019-10-22 03:04:52 -0500 )edit

That's what I was thinking, just something humans can easier read. That's why I asked if there was potentially something that let you get around it

stevemacenski gravatar image stevemacenski  ( 2019-10-22 14:02:16 -0500 )edit