/gym_ped_sim

A ros gazebo plugin for pedestrians (Raw depth social compliant navigation through GAIL) ICRA 2018

Primary LanguageC++BSD 2-Clause "Simplified" LicenseBSD-2-Clause

A ros gazebo plugin for pedestrians

This is a ros pkg for gazebo actor plugin.

Dependencies

  • Ubuntu 16.04
  • ROS-kinetic
  • Gazebo 8 (with actor suport)
  • python-lxml
  • turtlebot3
  • turtlebot3_msgs
  • turtlebot3_simulations

Build

  1. Add the repositories of Gazebo 8 and ROS kinetic

  2. Install Gazebo 8, Ros kinetic in buntu 16.04 and other dependencies.

sudo apt-get install ros-kinetic-desktop-full
sudo apt-get install ros-kinetic-gazebo8-ros-pkgs
  1. Build packages
cd /path/to/workspace/src
git clone git@github.com:onlytailei/gym_ped_sim.git
catkin build

rviz gazebo

Example

roslaunch turtlebot3_social default.launch

Node Details

  • actor_plugin
     Build based on a Gazebo official example. This node broadcasts the tf of every actor. Social force model is applied in every actor to interactive with each other.

  • actor_services
    The python files help to create several gazebo sdf files quickly. There is a rviz file for visualization.

  • turtlebo3_social
    In our socially compliant pedestrian simulator, we collect data by mounting a depth sensor onto one of the pedestrians, to the height matching that of real-world setups. Then, the social force model, as described in the paper, is used to label each incoming depth image with their corresponding social force.

  • data_collection
     To save the related dataset.

Dataset

The collected pedestrian navigation dataset contains:

  • depth image
  • RGB image
  • target
  • social force classification
  • social force
  • sum force

Interactive interface

Please reference gym_style_gazebo


This is the reference implementation of the plugins and for the paper Socially-compliant Navigation through Raw Depth Inputs with Generative Adversarial Imitation Learning. If it helps your research, please cite:

@inproceedings{tai2018social,
    author={L. Tai and J. Zhang and M. Liu and W. Burgard},
    booktitle={2018 IEEE International Conference on Robotics and Automation (ICRA)}, 
    title={Socially Compliant Navigation Through Raw Depth Inputs with Generative Adversarial Imitation Learning}, 
    year={2018}, 
    pages={1111-1117}, 
    doi={10.1109/ICRA.2018.8460968}, 
    ISSN={2577-087X}, 
    month={May},
}

References

srl-freiburg/pedsim_ros
social force model