This repository contains a ROSjava package enriching ROS with knowledge representation and reasoning services. The module has been specifically designed for Human-Robot Collaboration. The objective of the package is twofold: (i) the package proposes the use of standard semantic technologies (OWL/RDFS) based on an ontological framework for Human-Robot Collaboration [1] to represent and reason about production knowledge; (ii) the package implements knowledge reasoning and extraction mechanisms to facilitate the deployment of timeline-based task planning technologies [2] in HRC production scenarios.
The main components are ProductionKnowledge and ProductionKnowledgeAuthoring. The former encapsulates a Knowledge Graph representing information about HRC processes, collaborative tasks and human and robot capabilities/skills and, support reasoning and mechanisms. The latter support automatic generation of task planning models. It specificallly compiles the knowledge to dynamically generate a valid and updated timeline-based planning model [2] that can be used by tools like e.g., PLATINUm [3] to coordinate Human and Robot behaviors at task planning level [4,5].
The package has been developed and tested for ROS Melodic distribution on Ubunut 18.04. It requires a ROSJava workspace for correct execution and generation of the Java artifacts needed for custom messages and installed Java software.
We recommend the use of Java 8 since other higher or lower versions of Java may have some dependency issues with rosjava. Alternative versions of Java can be installed using the following command:
sudo update-alternatives --config java
In addition, the package requires an installed locally running instance of MongoDB server (the current package has been developed and tested using MongDB Cumminuty Server v.4.4.5).
Create an empty workspace.
mkdir -p ~/ws/src
cd ~/ws
catkin_make
Here are the instructions for downloading and building ROSJava from source (recommended). Please refer to the official ROSJava Wiki for further details about installation and availabe distributions.
wstool init -j4 ~/ws/src https://raw.githubusercontent.com/rosjava/rosjava/kinetic/rosjava.rosinstall
source /opt/ros/melodic/setup.bash
cd ~/ws
rosdep update
rosdep install --from-paths src -i -y
catkin_make
Note that one of the dependecies of rosjava_messages package is world_canvas_msgs which is not supported by ROS Melodic. To successfully build rosjava thus remove this dependency by editing the file package.xml of rosjava_messages and commenting the following line
<build_depend>world_canvas_msgs</build_depend>
Once that a ROSJava workspace has been successfully configured everything is ready for the installation of the sharework_knowledge package.
First, it is necessary to install the ROSjava package sharework_knowledge_msgs defining custom messages and services. This package is available on GitHub (https://github.com/pstlab/sharework_knowledge_msgs.git) and can be installed into the ROSJava workspace as follows:
cd ~/ws/src
git clone https://github.com/pstlab/sharework_knowledge_msgs.git
cd ~/ws
catkin_make
source ~/ws/devel/setup.bash
It is possible to verify the successful built of the package by checking ROS service description through rossrv show. For example the following commands
cd ~/ws
source devel/setup.bash
rossrv show sharework_knowledge_msgs/KnowledgeRDFUpdatePoint
should give the following output which describes the request and response defined for the service KnowledgeRDFUpdatePoint
string updateType
string[] data
---
sharework_knowledge_msgs/KnowledgeRDFTriple[] result
string subject
string property
string object
Before buliding the ROS package configure gradle so that all the dependencies are correctly downloaded. This module indeed relies on some packages (e.g., PLATINUm) that are not deployed on maven central but on GitHub.
To allow gradle to download packages from GitHub repositories create a text file gradle.properties under sharework_knowledge. This file should specifiy a GitHub username and a GitHub access token as follows:
gpr.user=<github-user>
gpr.token=<github-personal-access-token>
See how to create personal access token on the official GitHub page if necessary). Be sure to select the scope read:packages when creating the personal access token in order to successfully read GitHub packages.
As alternative to the file gradle.properties the same configuration parameters can be specified using environment variables. For example add to .bashrc the definition of the following variables:
export GITHUB_USER=<github-user>
export GITHUB_TOKEN=<github-personal-access-token>
At this point install the sharework_knowledge package into the ROSJava workspace by cloning and installing the current repository.
cd ~/ws/src
git clone https://github.com/pstlab/sharework_knowledge.git
cd ..
source devel/setup.bash
catkin_make
source devel/setup.bash
To finalize the installation just define the environment variable SHAREWORK_KNOWLEDGE in order to point to the folder containing the installed package.
export SHAREWORK_WS=<path-to-the-workspace>
export SHAREWORK_KNOWLEDGE=$SHAREWORK_WS/src/sharework_knowledge
The above line of code can be added to the .bashrc file to automatically export the environment variable when the terminal is open.
If the the installation and configuration of the sharework_knowledge package has been successfully done then it should be possible to run the ROS node starting the developed knowledge services.
First open a terminal and start core ROS nodes using roscore.
Then, open a different terminal and launch the service as follows:
cd ~/ws
source devel/setup.bash
rosrun sharework_knowledge production_knowledge it.cnr.istc.pst.sharework.service.KnowledgeService
At this point the service is running and can be called to interact with the knowledg base. As an example considered the following commands to be executed on a third terminal. The commands load one of the ontological models availabine under the folder etc/ontologies of hte package structure and query the model using respectively the /sharework/knowledge/update and /sharework/knwoledge/api services.
cd ~/ws
source devel/setup.bash
rosservice call /sharework/knowledge/update "load" ["<absolute_path-to-the-workspace>/src/sharework_knowledge/etc/ontologies/soho_cembre_v0.1.owl"]
rosservice call /sharework/knowledge/api "get_workers" []
The frist call load the ontological model defined for one of the case studies of the project in order to fill the knowledge base with the information characterizing the considered collaborative process. Please note that the file path specified into the "load request" should be absolute.
Note also that when the model is loaded the authoring service is triggered to automatically generate and validate an updated timeline-based planning model for the considered HRC scenario. The authoring process runs in background (i.e., it does not slow down queries/updates on the knowledge) and the output is the file prod_knowledge.ddl under the folder gen of the package structure.
The second call retrieves information about defined workers i.e., individuals of the ontological class SOHO:WorkOperator.
[1] Umbrico A., Orlandini A., Cesta A. "An Ontology for Human-Robot Collaboration". Procedia CIRP. Volume 93. 2020. Pages 1097-1102,
[2] Cialdea Mayer, M., Orlandini, A., Umbrico, A. "Planning and execution with flexible timelines: a formal account". Acta Informatica. Volume 53. 2016. Pages 649–680.
[3] Umbrico A., Cesta A., Cialdea Mayer M., Orlandini A. "PLATINUm: A New Framework for Planning and Acting". In: Esposito F., Basili R., Ferilli S., Lisi F. (eds) AIIA 2017 Advances in Artificial Intelligence. AIIA 2017. Lecture Notes in Computer Science, Volume 10640. 2017.
[4] M. Faroni, Beschi M., Ghidini S., Pedrocchi N., Umbrico A., Orlandini A., Cesta A. "A Layered Control Approach to Human-Aware Task and Motion Planning for Human-Robot Collaboration". 29th IEEE International Conference on Robot and Human Interactive Communication (RO-MAN). 2020. Pages 1204-1210.
[5] Pellegrinelli S., Orlandini A., Pedrocchi N., Umbrico A., Tolio T. "Motion planning and scheduling for human and industrial-robot collaboration". CIRP Annals. Volume 66, Issue 1. 2017. Pages 1-4.