Surrogate Assisted Generation of Human-Robot Interaction Scenarios

Code for the paper "Surrogate Assisted Generation of Human-Robot Interaction Scenarios".

Contents

• Manifest
• Installation
  • Notes
• Running Examples
  • Running Shared Control Teleoperation Example
  • Running Shared Workspace Collaboration Example
• Running QD Search Experiments
  • Running the Simulation Server
  • Running QD Search
• License

Manifest

• qd/: Code for QD search.
• src/: Code for simulating the scenarios.
• scripts/: Bash scripts for running experiments.

Installation

1. Install singularity.
2. Create the openrave container: sudo singularity build openrave_container.sif openrave_container.def. See openrave_container_README.md for more details about the container and potential issues.

Notes

1. When opening singularity shell or running commands in the container, set SINGULARITYENV_DISPLAY=$DISPLAY to allow GUI apps to work. Since X11 and local monitor use different environment variables, this needs to be set during execution time.
2. Use option --bind <path_to_this_repo>/src:/usr/project/catkin/src when opening singularity shell or running commands in the container. Otherwise, singularity's python will continue to import the original python files included during the build.

Running Examples

Note, you might require multiple singularity shells for this step. In the first shell, run roscore.

Run the python script in the second shell as described in the subsections below.

In the third shell, run rosrun rviz rviz.

In the rviz window, add MarkerArray with Marker Topic visualization_marker_array and InteractiveMarkers with Update Topic /openrave/update.

Running Shared Control Teleoperation Example

Commands to run the python script:

cd src/simple_environment/src/simple_environment
python scenario_generators/scenario_generator.py

In the rviz window, the robot should move towards a goal.

Running Shared Workspace Collaboration Example

Commands to run the python script:

cd src/simple_environment/src/simple_environment
python scenario_generators/collab_scenario_generator.py

In the rviz window, the human hand (visualized using a sphere) should move towards a goal and the robot arm should move towards a different goal.

Running QD Search Experiments

In this setup, we connect the QD code in this directory with the simulation code running on OpenRAVE. The simulation code is turned into a Python 2 Flask server which handles requests for OpenRave evaluations while this code uses pyribs on Python 3 and queries the server for evaluations.

Running the Simulation Server

Note that we assume Python 2.7.12, as that is what the machine uses.

Install the additional requirements:

pip install -r requirements_networking.txt

In one terminal, run:

roscore

In another, start the server with:

cd src/simple_environment/src/simple_environment
python search/server.py -c SERVER_CONFIG

Replace SERVER_CONFIG with search/config/experiment/experiment.tml for shared control teleoperation experiments or with search/config/experiment/experiment_collab.tml for shared workspace collaboration.

Running QD Search

See qd/README.md

License

This code is released under the MIT License. Some packages in src/ have a BSD-3 license.