Carla simulator & Autoware & Eagleye Configuration Issue

[Thetrafficlights]

Dear author, I encountered a problem when using carla in combination with autoware and eagleye. The positioning was inaccurate. I suspect it might be related to the data conversion of imu. Could you please take a look for me? The following is my configuration and terminal output:

terminal output:

comfig.yaml:

Trajectory：

[rsasaki0109]

gnss.velocity_source_topic must not be the vehicle speed but rather the velocity obtained from GNSS, expressed in the ECEF coordinate system.
https://github.com/MapIV/eagleye_private/blob/main-ros2/eagleye_util/gnss_converter/src/gnss_converter_node.cpp#L121-L123

[Thetrafficlights]

Thank you for your reply. Yes, I also discovered this problem later. Then I also want to ask about the direction of imu. When is it necessary for tf_converted_imu.cpp to transform the data of each axis of the imu, and what is the basis for the transformation?

[rsasaki0109]

This isn’t something the user really needs to worry about.
The issue is that the internal calculations in Eagleye aren’t performed in the ROS coordinate system, but with the default parameters it works fine when using ROS-coordinate inputs.

[Thetrafficlights]

All right. I have now successfully run carla's sensor data. However, I found that as the operation progressed, the error of eagleye/fix became increasingly large, and the positioning even showed discontinuity. I don't know how to position this problem now. Among them, I found that if the vehicle keeps driving in a counterclockwise circle, the heading value will become smaller and smaller without any limit. On the contrary, if the vehicle keeps driving in a clockwise circle, the heading value will keep increasing.

My topic on gnss speed is /odometry/velocity_enu. It is derived from the quaternion conversion of the vehicle body speed through the imu (the Y-axis direction of the quaternion always points north). Then /odometry/velocity_enu was converted by me to ecef_vel in the gnss_converter_node.

Clockwise

Counterclockwise

Some trajectory results (where sparse trajectories are true values) :

config.yaml:

tf_converted_imu.cpp: