Autonomous Initiation of Human Physical Assistance by a Humanoid

This repository contains the Finite State Machine (FSM) mc_rtc controller used in the experiments presented in our work titled "Autonomous Initiation of Human Physical Assistance by a Humanoid" published in RO-MAN 2020 conference proceedings

Author, Contact, BibTeX

Anastasia Bolotnikova (anastasia.bolotnikova[at]epfl.ch)

@inproceedings{bolotnikova2020roman,
  author = {Anastasia Bolotnikova and S{\'e}bastien Courtois and Abderrahmane Kheddar},
  title = {Autonomous Initiation of Human Physical Assistance by a Humanoid},
  booktitle = {IEEE International Conference on Robot and Human Interactive Communication},
  address = {Naples, Italy},
  month = {31 August--4 September},
  year = {2020}
}

Branches

• master - controller for real experiment using Azure Kinect Body Tracking visual feedback for navigation towards a person, and contact detection (DOI: 10.1109/ROMAN.2018.8525774, 10.1109/HUMANOIDS.2018.8624946) for establishing contacts with human in closed loop
• topic/withHumanModel - controller for experiment simulation with a human model

This repository is a fork of a base Pepper FSM mc_rtc controller PepperFSMController.

Required packages

• mc_rtc
• mc_pepper
• Azure_Kinect_ROS_Driver
• XGBoost
  • Note: do system install (e.g. set(CMAKE_INSTALL_PREFIX "/usr/local/")) to enable CMake to find_package(xgboost)

List of FSM states

• StandStraight
• NavigateToHuman
• IntentCommunication
• PreContactBack
• MakeContactBack
• PreContactHand
• MakeContactHand
• RemoveContacts
• MoveMobileBase

Cloning , bilding and installing the controller

git clone git@github.com:anastasiabolotnikova/autonomous_phri_init.git
cd pepperfsmcontroller-roman2020
mkdir build
cd build
cmake ..
make
sudo make install

Running the controller

Use your preferred interface to run this controller. Some option are:

• mc_rtc_ticker from mc_rtc_ros (simulation only)
• mc_naoqi (for experiments with real robot)

To run this controller, make sure to setup correctly following interface configuration entries:

# Main robot in controller
MainRobot: PepperExtraHardware
# Enabled controller
Enabled: PepperFSMController_RO-MAN2020

Visualizing the controller in RViz

To visualize controller state in RViz, you will first need to ensure that you have ROS installed and mc_rtc built with ROS support.

Run roslaunch launch/display.launch], using display.launch file provided in this repository, to launch controller display RViz node.

Once RViz node is started, configure it by clicking in RViz File -> Open Config -> select file roman2020_display.rviz, provided in launch folder in this repository

Add RViz configuration to ROS path

Optionally, you can configure RViz to use roman2020_display.rviz configuration automatically when running roslaunch launch/display.launch by adding it to your ROS path:

cp launch/roman2020_display.rviz ~/.ros/

Acknowledgment

Thanks to @gergondet and @arntanguy for reviewing and helping with this controller and the mc_rtc framework