robot_pose_ekf diverges with GPS data

asked 2014-09-19 03:15:28 -0500

mister_kay gravatar image


I use the viso2_ros visual odometry, IMU, wheel odometry and GPS data for robot_pose_ekf. By using VO, IMU and wheel I receive excellent results whereas when I also want to add my forwarded GPS-odometry topic (from navsat msg with gps_common->utm_odometry_node, from an xsens sensor) the filter diverges

I tried for some time now to manage this problem. As far I can see robot_pose_ekf interprets all topics as absolute measurements (using StampedTransform/Transform current minus last epoch). What looks weird ist that:

gps_common->utm_odometry_node sets the orientation to (quaternion): x = 1 y = 0 z =0 w = 0

which is weird since this is not the identity quaternion (x = 1 y = 0 z =0 w = 1 ). Also the x-axis bears the East-coordinates and not the North coordinates.

Has anybody experienced the same problems?

edit retag flag offensive close merge delete