Question Regarding Object Detection in Project

tjdtkadn · July 9, 2020, 2:06am

Hi,

Currently, I am working on Unit 6: Project.

This is my environment setting in simulation

I was trying to detect a cube with “basic_grasping_perception.cpp”.

I assumed I only need to modify name of subscribe topic for cloud data from the RGBD camera

This is what I did.

Then, I launched basic_perception_dbg.launch file.

And then I wanted to see if the robot detected the cube.

To do so, I did the following.

It said the process succeeded, but there is no detected object and surface.

Also, I realize that there are no camera-related nodes in this simulation when I check rqt_graph while there are many of camera-related nodes in fetch robot simulation at unit 5: Grasping.

e.g. rqt_graph for fetch robot in unit 5

e.g. rqt_graph for rb1 robot in unit 6

However, I do see the camera on my robot can see the objects( table and cube).

I don’t know what went wrong and what is going on here.

Can I get some help?

Also, I realize nobody answers my posts recently. I hope you guys take care of our posts more often.

Thanks.

albertoezquerro · July 9, 2020, 11:43am

Hello @tjdtkadn,

The problem is with the PointCloud of the simulation. If you visualize it on Rviz, you will see that it is intermitent (it keeps appearing and disappearing). This is causing the perception node to fail when the PointCloud disappears and to work normally when the PointCloud is there.

I don’t know why this is happening. I’m going to investigate more on this and see if I can come up with a fix.

Best,

tjdtkadn · July 10, 2020, 3:44am

Yes. I did see it’s happening. I thought it occurred because the cloud data is extensive.

Yes, please let me know when the simulation issue get fixed…

For, now I employed Forward Kinematic method assuming we are informed location of the object, but I do want to combine perception and motion planning.

Thank you for the response though.

Best,