Open AhmedTM opened 4 months ago
If ground_is_obstacle
is true, everything will be black (i.e., obstacles).
Yes thanks I noticed that and fixed it and I am still getting a very bad grid map I have added map_always_update
to true it got a little better but still bad when I use teleop at rotation the odometry fails and messes up the cloud map and the grid map. The below images are my best trials the first one is with lidar odomtery only I did a lot of rotations and the odometry was not bad the second one was with robot_localization ekf and I didn't make any turns. Do you have any advice to improve the result.
For issues with robot_localization
, you may ask on their project page. If odometry produced by icp_odometry is not akay, then maybe I can help. If you have a recording of the lidar messages points
, I could take a look to see at least if icp_odometry is working as expected.
For the grid, is there an issue with position error or that the empty space is not filled enough with empty cells?
Sorry for not being clear I have stopped robot_localization
package and I was using icp_odometry
only at this tests I also added Grid/MaxGroundHeight
and Grid/MaxObstacleHeight
parameters for rtabmap_slam
to get a better grid and the result is better but I still have an issue with the position. Sometimes it feels like the map is not fixed although I am using map
as the fixed frame in rviz. Here is a bag file recorded from the lidar data I am using a gazebo simulation environment the lidar is HDL32E and I limited its range to 180 degree because the back of the cloud have some problems. So, you will only find the 180 degree range of the cloud in the bag file. If you want me to record a full bag with the camera and odometry and the lidar please tell me. Appreciate your help.
lidar bag
After 27-28 sec into the bag, there is no geometry complexity, all points are on the same plane:
icp_odometry
will think the robot is not moving during that time or may do very bad drift.
Yes, I can see it there is a part in the road where there is not geometric complexity in the FOV of the LiDAR sensor since it's an outdoor environment. That's why I was trying to fuse wheel odometry with LiDAR to somehow fix that but now I can see a new node in my rtabmap build that is called rgbdicp_odom
but I can't find its documentation could this kind of fusion fix this issue ?
Can you give a screenshot of the simulated environment?
Here are some common challenges that simulated environments can have:
If your simulated environment checks two or more of them at the same time, testing visual and/or lidar SLAM approaches may be even more difficult in the simulator than in reality. For example, if you have 1 and 2 at the same time, visual odometry won't help, lidar odometry won't help, thus you would need to rely only on wheel odometry. In your case I wouldn't use rgbdicp_odometry
but try rgbd_odometry and icp_odometry in parallel to your wheel odometry. As you are targeting outdoor environments, you may try something like this (assuming wheel odometry is publishing on odom
frame):
rgbd_odometry
:
guess_frame_id=odom
odom_frame_id=vo
Odom/ResetCountdown=1
icp_odometry
:
guess_frame_id=vo
odom_frame_id=icp_odom
Odom/ResetCountdown=1
You may need to adjust some ICP parameters to avoid icp_odometry to jump to out of the input guess. Also remap output odometry topics. Using guess_frame_id
like this, when one of the odometry is lost, it will still republish the input guess on TF so that the following TF tree is never broken:
icp_odom -> vo -> odom -> base_link
Hello guys, First I would like to thank you on this awesome work your package is really helpful and easy to use and debug. I am trying to build a gridmap for an outdoor environment simulation on gazebo I am using HDL32E LiDAR, D435 camera and wheel odometry this is how my launch file.
The point cloud map is correct and the localization path is correct but the grid map is not correct. This is a screenshot of the output gridmap on rviz:![Screenshot from 2024-02-23 04-02-43](https://github.com/introlab/rtabmap_ros/assets/22213427/65be4e11-1dc0-484b-bc7f-969c0eedd134)
I know for fact I am setting some parameters wrong can you help me with this.