This is a ROS package that represents a risk aware decision making framework for assistive robots. It is authored by Roya Sabbagh Novin, Amir Yazdani, Andrew Merryweather, and Tucker Hermans from Utah Robotics Center, University of Utah.
More information about the paper and the method can be found in the paper website and a PDF version is available at the arXiv pape.
- Functionality:
- Risk-aware decision making for an assistive robot
- Find the best intervention plan
- Considering risk of fall for the patient
- leveraging CVaR risk metric
- Human motion/intention prediction and fall score assessment
- Generated human motion data in hospital room
- Learned GPs for human motion
- Predict probability distributions over intention and trajectory
- Using developed fall score assessment model (Github repository for fall risk evaluation)
- Simulation
- Gazebo visualization for simulation results
- Provided Gazebo actor plugin to simulate patient motion
- Including urdf for common objects in a hospital room
- Risk-aware decision making for an assistive robot
- Input: Robot intial position, object type, object initial position and orientation, patient initial position and orientation, layot of the environment
- Output: Human motion prediction, best intervention plan for robot, executed path for robot
- Operating System: Ubuntu (16.04), ROS kinetic
- Gazebo 8+ Install Gazebo any version above 8 should work. This package has been tested with Gazebo 9.12.0 for ubuntu.
- youBot packages Install youbot_simulation and youbot_description packages. (link)
- Make sure you have ROS-kinetic installed and your catkin workspace is generated
- Clone the package in your catkin workspace
- Make/build your catkin workspace
- Enjoy!
- Set a simulation example in "main_simulation.py" or choose one from the available ones"
- Run "launch_gazebo.launch"
Please cite these papers in your publications if it helps your research.
@article{novin2020risk,
title={Risk-Aware Decision Making in Service Robots to Minimize Risk of Patient Falls in Hospitals},
author={Sabbagh Novin, Roya and Yazdani, Amir and Merryweather, Andrew and Hermans, Tucker},
journal={arXiv preprint arXiv:2010.08124},
year={2020}
}
Links to the papers: