About tf tree

[AriYu]

I checked tf tree for adding /odom frame and found strange transforms like follow image. I added hokuyo3d_link before checking tf tree. I think base_footprint should be transformed from base_link or nearly root of tree. Please tell me why it is, thank you.

[RyodoTanaka]

@AriYu Year, I understand your saying. And, I think It is not good (wrong) statement. But there is a reason why I set such a strange tf tree. I'll tell you the situation.

First, I set the base_footprint be transformed from base_link. Then, I simulated with Gazebo. And I found the base_link was moving. That means base_footprint was moving. I thought base_footprint should be on the ground and not move (to Z axis direction). So, I set these strange tf trees.

Of course I know it's not good idea, or may be I think, we don't need to do that. But I have no idea. Please let me know your idea.

[MoriKen254]

I'm not familiar with the issue but let me give something here.

have a look at TF tree of p.310 of Learning ROS for Robotics Programming 2nd Edition. all related files about the tree are also available on Github including xacro of a very simple wheel robot.

hope u guys will find some hints from that.

[AriYu]

@RyodoTanaka OK, I understood. But, why did you try to transform base_footprint from base_link ? I read Learning ROS for Robotics Programming 2nd Edition, the book say base_footprint => base_link => wheels is better.

@mum254 Thank you for your suggestion.

[RyodoTanaka]

@AriYu Sorry, I had mistake.

First, I set the base_footprint be transformed from base_link.

That's not correct. I wanted to say

I set the base_link be transformed from base_footprint.

I mean, I set the tree base_footprint -> base_link. If you want to check that, go 72001a6bd18ebd8427af832ac2ffb376a0ec905f. And on the next commit (c9a5be3752d1d337ee8d7d0097c058869a3e1279), I changed that to now we use.

[RyodoTanaka]

@AriYu Now, we don't use the gear on the middle of the robot. Because, we need more resource for calculation and the model would be more complex to simulate. But, we have never tried that. And, we have another problem on the tf tree. So now, It worth to add the gear on the simulated model I think. Even if we need to recreate some stl files... Please give me your opinion.

[AriYu]

@RyodoTanaka OK, I see. For the time being, I will go under this structure of tf tree. If we found better solution, fix this problem. So, this issue will not close until this problem is solved.

[AriYu]

In fix-tf-tree branch, we tried new structure of tf tree fr01. A new structure of tf tree is like this. base_footprint => base_stabilized => base_link => other_links I referenced hector_slam tutorial. That link is here. And, this structure looks no problems. fr01_modified_TF_tree.pdf But, the modify will influence other programs such as using fake_base_footprint, because fake_base_footprint is deleted. Then, some other problems maybe occur. So, I want to get your agreement. Thank you @RyodoTanaka