robot_localization with kitti_dataset
Hey guys,
i am using the kitti dataset to test some localization with the named package. I got the following problem:
One node is calculating the pose from the IMU data and another one is calculating the pose from the GPS data.
I am looking at the poses in rviz with the images/video to check the trajectory. The IMU is working very well.
Unfortunately the GPS is weird and kind of non-deterministic. The .bag-File is continously firing GPS Signals which are not equal to zero. FYI I am really talking about playing the same sequence/bag-File. In some plays the Navsat-Odom is quite good. In some plays the Navsat-Odom jumps sometimes to zero, although the bag-file is sending a GPS value which should result in non-zero. In some plays the Navsat-Odom Signal gets so weird that it is the most of the time zero and sometimes jumps to the correct value.
Do anybody could help me? I do not understand why. I've also used other datasets and that problem doesnt occure. The Navsat-Config is as follows:
param name="frequency" value="100"/>
param name="magnetic_declination_radians" value="0"/>
param name="yaw_offset" value="0"/>
remap from="/imu/data" to="/kitti/oxts/imu" />
remap from="/gps/fix" to="/kitti/oxts/gps/fix" />
Should I add some more params to avoid that case?
Id be very glad for your help.
Bryan
There are too many variables with GPS data, and it can usually jump from maybe a meter to hundred meter depending on your position and surrounding. But the behavior you explained is a bit weird, maybe something is wrong with GPS data in that bag file in question. Can you tell me which node are you using to do calculations for you?
I think the .bag-File should be correct (the GPS output of kitti is stable), i got that phenomenon in all .bag Files of the kitti-dataset..I am using the Navsat-Node for the GPS and the EKF-Node for the IMU...Both Nodes are part of the robot_localization package..Appreciate your help