Only released in EOL distros:   hydro

This package wraps a Human-Robot Interaction Experiment conducted at the Univesity of Texas At Austin. The results from this experiment were reported in the following symposium paper: Piyush Khandelwal and Peter Stone. Multi-robot Human Guidance using Topological Graphs. In AAAI Spring 2014 Symposium on Qualitative Representations for Robots (AAAI-SSS), March 2014. (pdf)

The bwi_guidance_concert package wraps this experiment to use the Robotics in Concert framework.

Overview

The goal of the multi-robot guidance project is to allow an automated system to efficiently guide a person unfamiliar with the environment to their goal using multiple robots. In the experiment available through this package, a single controller decides where to position robots in the environment and what direction should be displayed on each robot's screen to guide the person.

This package currently wraps preliminary work on this project, where a central positioner node teleports each robot to a desired location using Gazebo's set_model_state service. Additionally, each robot subscribes to an image stream which is displayed on its simulated robot screen (ex. '/robot1/image'), and these images are also published by the positioner.

Apart from the positioner node, an experiment controller node decides which exact problem the user is facing, and a server node keeps track of which user is interfacing with the system. A user controls a human avatar inside the simulator using a web based GUI that talks to the rest of the system using rosbridge_suite.

The workflow of the system is as follows:

• Initially, we start a concert with the server node which accepts new users using the web interface.

• When a new user arrives and locks the server (preventing other users from connecting to the experiment until this lock is relinquished), the server internally starts gazebo, the positioner node and the experiment controller node within the service. In order to spawn all the robots, a wrapper around concert_service_segbot_gazebo is used, although all of these modules are loaded using roslaunch.

• 10 concert clients are loaded (1 for each robot). No topics are flipped to these concerts since these topics eventually need to be used locally, and without topic remapping and subscribers behaving poorly when subscribed to multiple masters (#127), it is typically a bad idea to flip these topics. Please note that when we move to the next version of the research, where robots use the standard navigation stack to navigate into place, none of the topics flipped will be required locally and the system can work as intended.

• A dummy app is run on each concert (empty launch and interface file), just to demonstrate how robots will be used in the next stage of research.
• At this time, the user runs through a series of experiments, data is collected and compiled into results.

Installation

• Follow the instructions at segbot_gazebo_concert#Installation to install all the base code required by this package.

• Additionally, you'll need to install a web server. This package has only been tested using the apache web server:

  sudo apt-get install apache2

• Next, install the following rosinstall entry. You may already have some of these packages if you've already installed the segbot_gazebo_concert demo. A patched version of rosbridge is used here (#104)

  - git: {local-name: segbot_rocon, uri: 'https://github.com/utexas-bwi/segbot_rocon.git', version: master}
  - git: {local-name: bwi_guidance, uri: 'https://github.com/utexas-bwi/bwi_guidance.git', version: master}
  - git: {local-name: rl_pursuit, uri: 'https://github.com/utexas-bwi/rl_pursuit.git', version: master}
  - git: {local-name: rosbridge_suite, uri: 'https://github.com/piyushk/rosbridge_suite.git', version: patch-1}

• You'll need to copy/symlink the necessary files so that they can be displayed by the webserver. If you've got home directory encryption enabled, then the following command won't work and you'll have to copy the files. The following command also assumes standard installation with the web home directory being /var/www.

  sudo ln -sf `rospack find bwi_guidance`/www /var/www/exp1

• Finally, you'll need to copy some data files to the bwi_guidance package which are required by the experiment:

  roscd bwi_guidance_solver/data/exp2/cache
  wget http://cs.utexas.edu/~piyushk/share/vi.tar.gz
  tar xvzf vi.tar.gz
  rm vi.tar.gz

Running the code

Command Line Execution

The following commands need to be run on a reasonably powerful machine (Recommended: Quad Core i7 Processor) that can also run Gazebo.

Non-Concert Version

roslaunch bwi_guidance_solver server.launch --screen

Concert Version

rocon_launch bwi_guidance_concert guidance.concert --screen

Running the User Interface

The user interface can be run on any browser that can connect to the machine on which the software was run. The connection is made using the rosbridge library over websockets.

Typically, you'll have to type the following link into the web-browser to launch the user side GUI. This link will only work on the machine from where you ran the command line application.

http://localhost/exp1/index.html

If you want to run on a different machine, you'll have to supply the hostname/IP of the machine where the command line application ran, so that your browser knows to connect to it:

http://robot-devil.csres.utexas.edu/exp1/index.html?host=robot-devil.csres.utexas.edu

Additionally, if you're running the concert version of the demo, you'll have to supply an additional parameters:

http://robot-devil.csres.utexas.edu/exp1/index.html?host=robot-devil.csres.utexas.edu&concert=true

A convenience webpage where you can enter this information and get redirected to the correct link is available at:

http://robot-devil.csres.utexas.edu/exp1/link.html

Once you get to the index.html page, follow the instructions to run through the experiment!