Revision history [back]

I am currently dealing with the same issue. When the rgbd_odometry node loses the odometry, EKF cannot understand it and it tries to predict odometry using the state transition model without measurement data (in case of IMU has not been used). Therefore the generated map starts to distort, from the moment the first odometry loss happened. There should be a way to communicate this odometry losing moment to the EKF, so it can stop predicting inaccurate odometry values. That is what happens when we feed visual odometry generated by rgbd_odometry node to the rtabmap node. rtabmap node identifies when odometry is lost, and it stops mapping then.

I am currently dealing with the same issue. When the rgbd_odometry node loses the odometry, EKF cannot understand it and it tries to predict odometry using the state transition model without measurement data (in case of IMU has not been used). Therefore the generated map starts to distort, from the moment the first odometry loss happened. There should be a way to communicate this odometry losing moment to the EKF, so it can stop predicting inaccurate odometry values. That is what happens when we feed visual odometry generated by rgbd_odometry node to the rtabmap node. rtabmap node identifies when odometry is lost, and it stops mapping then. until odometry is properly given.

I am currently dealing with the same issue. When the rgbd_odometry node loses the odometry, EKF cannot understand it and it tries to predict odometry using the state transition model without measurement data (in case of IMU has not been used). Therefore the generated map starts to distort, from the moment the first odometry loss happened. There should be a way to communicate this odometry losing moment to the EKF, so it can stop predicting inaccurate odometry values. That is what happens when we feed visual odometry generated by rgbd_odometry node to the rtabmap node. node directly. rtabmap node identifies when odometry is lost, and it stops mapping until odometry is properly given.