MSF localization thinks the vehicle is always slowly spinning

[josh-wende]

Hi,

I'm trying to get MSF localization running with LGSVL simulator. I made a local map of their CubeTown level, but the localization results are generally way off. I have attached a video that shows this behavior, you can see that once the vehicle is remaining stationary, the fusion localization result hones in on the correct position (but the orientation is still off due to the spinning). I have tried to figure out the source of the erroneous angular velocity, but have been unable to track it down. Do you have any ideas what might be causing this?

The possible sources of the issue that I have thought of are:

1. Nothing is published to the /apollo/sensor/gnss/heading topic, which the MSF localization component subscribes to.
2. Could it be an issue with the extrinsic calibration of the lidar on the vehicle?
3. The sins functions, but these are not open source. I don't know how it tries to remove bias from the sensor readings.

Any insight would be appreciated. Thanks in advance for the help!

https://user-images.githubusercontent.com/93289658/154392378-91cc585e-e1d4-4631-941f-71c28def8be3.mp4

[daohu527]

How did you create the local map and visualization, actually this part is not open source.

[josh-wende]

Hi @daohu527, I used the msf_simple_map_creator.sh script to create the local map from a cyber record recorded in LGSVL. I didn't record too much data, so the resulting local map is a bit fuzzy, but that doesn't seem like the issue here.

[daohu527]

Nothing is published to the /apollo/sensor/gnss/heading topic, which the MSF localization component subscribes to.

It depends on the configuration of the sensor, e.g. lgsvl

Could it be an issue with the extrinsic calibration of the lidar on the vehicle?

I'm not sure, if it is a simulation, your calibration is theoretically completely accurate

[chenhuanguang]

Closed due to inactivity. If the problem persists, pls feel free to reopen it or create a new one and refer to it.