Pose ekf slam topic mapping.
Hi all,
I am trying to use poseekfslam package as is given here.
The nodes that the method expects as are described in the package are the following:
Position fixes of type PoseWithCovarianceStamped in the topic /poseekfslam/pose_update
Linear velocity updates of type TwistWithCovarianceStamped are received through the topic /poseekfslam/velocity_update
Vehicle orientation and angular velocity are expected at /poseekfslam/imu_input
using an Imu sensor message. Finally, landmark updates have to be published at /poseekfslam/landmarkupdate/landmarkid
I am currently having an IMU sensor giving messeges at topic /odomimucombined which is a message of type Odometry.
The mapping I have done now is similar-ish to that but I was wondering if someone has had experience with the package and could validate that it's ok.
I have added a listener to /odomimucombined and mapped it to the PubPose[line 93] function of the package
I have added a listener to /mobilse_base/commands/velocity and mapped it to pubVel[line 67] function
Finally the imu input is also mapped to /odomimucombined as it contains imu information.
My questions are the following:
How are the messages in pubImu, pubVel, pubImu related to the transformation after the message publishing(lines 61,87,115 in testekfposeslam)? Should I modify something there? Also, I take mu odometry from /imucombined which is part of /poseekf node and it already contains integrated imu information. Is it wrong? Will the imu transformations to the data be applied twice to my pose due to that? If so, do I have to use my /odom topic instead? I don't know if it makes sense, if you need some more specifications regarding my current setup, please let me know.
Cheers,
Panos
Asked by panos on 2015-05-15 06:50:41 UTC
Comments