Revision history [back]

If you haven't already, you might want to check out robot_pose_ekf which does pretty much what you're describing: combining odom and imu data to output a transform that is better than either sensor taken separately.

If you're set on making your own version of robot_pose_ekf--the first thing I would do is visualize the odom and imu rotation vectors separately (and simultaneously) in rviz to make sure your IMU is outputting correct rotations with respect to the odom.

Here's a simple node that inputs an Imu message and outputs a Pose message (for visual debugging)

If you haven't already, you might want to check out robot_pose_ekf which does pretty much what you're describing: combining odom and imu imu/gyro data to output a transform that is better than either sensor taken separately.

If you're set on making your own version of robot_pose_ekf--the first thing I would do is visualize the odom and imu rotation vectors separately (and simultaneously) in rviz to make sure your IMU is outputting correct rotations with respect to the odom.

Here's a simple node that inputs an Imu message and outputs a Pose message (for visual debugging)