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

3. Build packages

cd /path/to/workspace/src
git clone git@github.com:onlytailei/gym_ped_sim.git
catkin build

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:

@article{tai2017socially,
  title={Socially-compliant Navigation through Raw Depth Inputs with Generative Adversarial Imitation Learning},
  author={Tai, Lei and Zhang, Jingwei and Liu, Ming and Burgard, Wolfram},
  journal={arXiv preprint arXiv:1710.02543},
  year={2017}
}