This package provides a user interface for collecting workspace trajectories using HERB's end effector.

You need to install the rosbridge_suite. Rosbridge server is part of the rosbridge_suite of packages, providing a WebSocket transport layer. By providing a WebSocket connection, rosbridge server allows webpages to send and recieve ros messages and service calls using the rosbridge protocol.

cmd> sudo apt-get install ros-indigo-rosbridge-server

One component the query is comprised of is env: the OpenRAVE environment. These queries are scenes that we want the users to complete. To execute queries the following additional installation steps are necessary:

cmd> apt-get install ros-indigo-gperftools-21

Then checkout randomized_rearrangement_planning, or_fcl and the branch feature/new_planning_benchmarks of repository benchmarks into your catkin workspace.

A variety of queries are availabe in the queries/ directory. To make a new query, execute an existing query and you can add, remove and move kinbodies. To make a new query:

$ cd tools
$ python make_query.py --query ../queries/bh280_1movable.query --savefile bh280_temp.yaml \
--viewer interactivemarker

After execution an IPython console will have opened from which you can add or remove kinbodies. Open rviz, add interactivemarker and set Update Topic to /openrave/update. To move objects around right-click on the object, hover over Body and click Pose Controls. You can now drag the objects around. When have finished making the query, exit out of the IPython console and your new query will be saved to the file you passed to the savefile argument.

There a few things that need to be changed for you to be able to deploy the study.

First, if you would like to use a different set of queries than those already in the queries/ directory you will need to change the paths in the queryfiles list in user_interface_web/run_back_end.py.

Next, in user_interface_web/site.py change /home/ubuntu/results/ in line 65 to the directory you would like to save the results. Change app.run(host='0.0.0.0', port="YOUR_PORT_NUMBER") in line 74.

If you changed the number of queries in user_interface_web/run_back_end.py change if req.sceneNum < NUM_QUERIES: in line 96 of user_interface_web/state_updater.py. Finally, change /home/ubuntu/results/ in line 86 and 105 to the directory you would like to save the results.

To the deploy the study on your local computer, first launch the rosbridge_server in a new screen and then run site.py:

$ screen -S rosbridge
$ roslaunch rosbridge_server rosbridge_websocket.launch port:=8888
$ <CTRL+a><d> #detach from the screen
$ python site.py

To view the study, type http://localhost:YOUR_PORT_NUMBER/ in the browser.

Before you can deploy this study on Mechanical Turk, you need to create an EC2-Instance on Amazon Web Services. See Clint Liddick for details.

Often it is desirable to render images outside of the OpenRAVE environment. There exist a suite of tools for doing this for tests involving the bh280.

To render a path:

$ python play_path.py --query results/query.yaml --path results/path.ompl

Note this will simply provide snapshots of each point in the path. The original path resolution can be quite course. To subsample the path in order to enable playback at a finer resolutions:

$ python subsample_paths.py --dir results/
$ python play_path.py --query results/query.yaml --path results/path_subsample.ompl

To generate a video of the playback:

$ python play_path.py --query results/query.yaml \
--path results/path_subsample.ompl --save --outdir results/frames/
$ python make_video.py --image-dir results/frames/ --outfile results/video.mp4

The following helper script can be used to subsample, playback and generate a video all at once:

$ python make_path_video.py --query results/query.yaml \
--path results/path.ompl --outfile video.mp4

user_interface is licensed under a BSD license. See LICENSE for more information.

user_interface is developed by Vinitha Ranganeni (@vinitha910) at the Personal Robotics Lab in the Robotics Institute at Carnegie Mellon University. I would like to thank Clint Liddick (@ClintLiddick) and Jennifer King (@jeking) for their assistance in developing user_interface.