cmu-tbd / hri18-bystander-intervention-study

Bystander Intervention Study

COPYRIGHT(C) 2019 - CMU Assitive Robots Lab - Code released under MIT.
Contact - Zhi - [email protected]

Update: This branch (replication) consist of code we are working on to be used in upcoming replication studies. For the original code used in the 2018 paper, please look at the Master branch.

This repository contains the code that ran the user study described in Inducing Bystander Interventions During Robot Abuse with Social Mechanisms by Tan et al. (2018). Due to the experimental nature of the work, the code does not meet industry coding guidelines / best practices and would need modifications to work on different machines. We provided a brief guide and pointers on how the system work and requires hardwares.

Hardware

• ANKI Cozmo Robot
• Smartphone that can run Cozmo App (We used an Iphone 5C in the original study)
• Tablet (Samsung Galaxy Tab 10.1 in the original study)
• Micosoft Kinect 2
• Laptop with Ubuntu 16.04 (i7 Lenova laptop in the original study)

Software/ROS Dependencies

• ROS Kinetic
• Nodejs
• CMU-ARM/alloy
• SUCCESS-MURI/success_google_stt
• SUCCESS-MURI/success_ros_msgs
• SUCCESS-MURI/success_ros_aruco
• iai_kinect2
• audio_common
• our heavily modified cozmo_driver

You will also need a google cloud account that has speech to text API enabled.

Installation

1. Install ROS and all the dependencies above.
   • We included a .rosintall file that simplifies the process, wstool update study.rosinstall
2. Install all python packages listed in the requirement.txt
3. Download this ROS package.
4. Install Nodejs
5. Install all the requirements of Nodejs located in learning_server/package.json
   • at learning_server run npm install
6. Change the var ros_url' in learning_server/public/js/common.jsto the IP address or URL of the machine running the code.

Basic Instructions on Running

1. use roslaunch to start the base launch file. roslaunch bystander_intervention_study base.launch
   • This starts the kinect, rosbridge and other dependencies.
2. use rosparam to set the conditions
   • the parameter bully_reaction is a string that sets to the types of responses:none,shutdown and emotional.
   • the parameter empathy_flag is a bool that when set to true makes the empathetic during the task
3. use roslaunch to start the cozmo launch file. roslaunch bystander_intervention_study cozmo.launch
   • This starts the robot.
4. Start the nodejs server that runs the learning application with nodejs learning_server/server.js
5. start the study by running the python script, study.py with rosrun bystander_intervention_study study.py
6. You should be able to connect to the learning application at 128.0.0.1:3000 on your local machine.

We included a startup script,start.sh that we used in our original study to start the program. There are several hard links in the script that will need to be modified to work correctly.