search

ZJU-FAST-Lab / ego-planner

GNU General Public License v3.0
1.31k stars 269 forks source link

Question about airsim simulation #38

Closed rsj007 closed 2 years ago

rsj007 commented 2 years ago

Hello, I am running ego-planner simulation in airsim. I use depth and camera pose as inputs .The tfs I set as follows. However the grid map is strange, only a small one, not like the tree before the drone. Besides when the drone take offs, the grid map disappears. Could someone give me some guides please?

Screenshot from 2022-03-24 19-07-04

bigsuperZZZX commented 2 years ago

Obviously, the extrinsic parameters of the camera are inappropriate. Check the framework of your camera pose topic, we don't know what frame airsim is used. Then refer to this line of code to write your own code for rotating your acquired camera pose.

rsj007 commented 2 years ago

I use the point cloud as input. The tfs I set are as below: <node pkg="tf" type="static_transform_publisher" name="camera_link_to_world" args="0 0 0 -1.57 0 -1.57 base_link front_center_optical 30" /> <node pkg="tf" type="static_transform_publisher" name="tf_camera" args="0 0 0.0 0 0 0 base_link odom 100"/>

Is it right? Do I need other tfs? Thanks very much!

LvAoAo commented 2 years ago

I use the point cloud as input. The tfs I set are as below: <node pkg="tf" type="static_transform_publisher" name="camera_link_to_world" args="0 0 0 -1.57 0 -1.57 base_link front_center_optical 30" /> <node pkg="tf" type="static_transform_publisher" name="tf_camera" args="0 0 0.0 0 0 0 base_link odom 100"/>

Hello! I may meet the same question... So could you please share something about how to get the correct depth image required by the "ego-planner"?I input the depthplanar and just get 2-value images just like the photo.....And I nearly get no grid map any time.Only when the drone is encountered with an obstacle can i got a slight gridmap... grid

bigsuperZZZX commented 2 years ago

Our mapping code does not use tf to get extrinsic parameters. You should manually set it directly in the code as I mentioned in the previous response.

WjMessi1 commented 2 years ago

I use the point cloud as input. The tfs I set are as below: <node pkg="tf" type="static_transform_publisher" name="camera_link_to_world" args="0 0 0 -1.57 0 -1.57 base_link front_center_optical 30" /> <node pkg="tf" type="static_transform_publisher" name="tf_camera" args="0 0 0.0 0 0 0 base_link odom 100"/>

Hello! I may meet the same question... So could you please share something about how to get the correct depth image required by the "ego-planner"?I input the depthplanar and just get 2-value images just like the photo.....And I nearly get no grid map any time.Only when the drone is encountered with an obstacle can i got a slight gridmap... grid

So could you get the right depth image now?

bigsuperZZZX commented 2 years ago

If there is no more reply, I will close this issue