search

jsk-ros-pkg / jsk_control

jsk control ros packages
http://github.com/jsk-ros-pkg/jsk_control
13 stars 51 forks source link

Menu based teleop plugins #721

Open weiqiyang opened 5 years ago

weiqiyang commented 5 years ago

This PR adds several plugins to jsk_teleop_joy that provides functions including base movement, bounding-box selection and end-pose adjustment.

Following is the demo video. https://drive.google.com/file/d/1WrQpt2VzYSj-bA00aqcJO4Gx3hJAHAHz/view?usp=sharing

knorth55 commented 5 years ago

Can you attach demo video about how to use this plugin?

weiqiyang commented 5 years ago

I add a sample launch file. There are still several improves that I would like to imply, so I add WIP to the title.

weiqiyang commented 4 years ago

I modified the title and PR comment with a link to a demo video.

k-okada commented 4 years ago

Nice video

  1. What is the role of irteus view?
  2. I think latency is very important in this application, Please analyze why the point loud did has latency + low throughput . For example, if we subsmple the cloud, can we speed up? Or did you transfer pointcloud of depthmap ...
weiqiyang commented 4 years ago

Thank you.

1. What is the role of irteus view?

The plugins publishs manually adjusted pose and button commands. The irteus view in the demo is an example of the application of the plugins cooperating with euslisp.

2. I think latency is very important in this application, Please analyze why the point cloud did has latency + low throughput . For example, if we subsample the cloud, can we speed up? Or did you transfer pointcloud of depthmap ...

The pointcloud has latency because I was using the raw data while making this video. We tried using compressed data republished by an extra node later, which do speed up the pointcloud display.

k-okada commented 4 years ago

If operator only sees Rviz, then I think we do not need irtview in the video. If you use pointlcoud, please consider using depthimage

https://github.com/jsk-ros-pkg/jsk_robot/issues/923

2019年11月1日(金) 16:49 Weiqi Yang notifications@github.com:

Thank you.

  1. What is the role of irteus view?

The plugins publishs manually adjusted pose and button commands. The irteus view in the demo is an example of the application of the plugins cooperating with euslisp.

  1. I think latency is very important in this application, Please analyze why the point cloud did has latency + low throughput . For example, if we subsample the cloud, can we speed up? Or did you transfer pointcloud of depthmap ...

The pointcloud has latency because I was using the raw data while making this video. We tried using compressed data republished by an extra node later, which do speed up the pointcloud display.

— You are receiving this because you commented.

Reply to this email directly, view it on GitHub https://github.com/jsk-ros-pkg/jsk_control/pull/721?email_source=notifications&email_token=AADYNXCVDS2LXZNYSLKYLZ3QRPNP7A5CNFSM4HY64MQKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEC2HRKQ#issuecomment-548698282, or unsubscribe https://github.com/notifications/unsubscribe-auth/AADYNXD5KWZYTJIMQ4IKYMLQRPNP7ANCNFSM4HY64MQA .

--

◉ Kei Okada

weiqiyang commented 4 years ago

If operator only sees Rviz, then I think we do not need irtview in the video.

The operator sees the irtviewer also, to confirm ik result before executing. Maybe if we can have an euslisp script that generate a tf topic from joint-angle array, then we can show the result as another robot model, and the irtviewer is no longer needed.

If you use pointlcoud, please consider using depthimage jsk-ros-pkg/jsk_robot#923

Yes, this is what I mean by "compressed data", and we are using this node now.

k-okada commented 4 years ago

You can also use https://github.com/jsk-ros-pkg/jsk_roseus/blob/master/roseus_tutorials/src/robot-publisher.l

If you’d like to pursuit UI research , user experience and completeness is very important

If you already using image depth , please analyze this it is so slow

2019年11月1日(金) 17:23 Weiqi Yang notifications@github.com:

If operator only sees Rviz, then I think we do not need irtview in the video.

The operator sees the irtviewer also, to confirm ik result before executing. Maybe if we can have an euslisp script that generate a tf topic from joint-angle array, then we can show the result as another robot model, and the irtviewer is no longer needed.

If you use pointlcoud, please consider using depthimage jsk-ros-pkg/jsk_robot#923 https://github.com/jsk-ros-pkg/jsk_robot/issues/923

Yes, this is what I mean by "compressed data", and we are using this node now.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/jsk-ros-pkg/jsk_control/pull/721?email_source=notifications&email_token=AADYNXEYNJI7I3HXA3LEMGDQRPRNJA5CNFSM4HY64MQKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEC2JPSA#issuecomment-548706248, or unsubscribe https://github.com/notifications/unsubscribe-auth/AADYNXGKQRAQXLICFAJLQXLQRPRNJANCNFSM4HY64MQA .

--

◉ Kei Okada

weiqiyang commented 4 years ago

Thank you for the suggestion!

I tried to modified the code to apply it on hsr and successfully published the marker. However, since the marker topic is presented in points, the topic was very heavy. (rostopic hz showed a result of less than 0.2Hz. I also tried to take a log by rostopic echo -n 1, and a single publish was over 15MB)

Now I am using robot_state_publisher. I will try to come up with a new video this week.

knorth55 commented 4 years ago

Can you use 4 spaces for indentation? Python does not matter, but it is better to use same indentation as other python files.

knorth55 commented 4 years ago

There are many changes, so it may take time for review. I'm reviewing now.

weiqiyang commented 4 years ago

Can you use 4 spaces for indentation? Python does not matter, but it is better to use same indentation as other python files.

I checked other python files in the plugin folder, and I think they are all using 2 spaces for indentation.