This repository contains sample code and resources to help with the implementation of the policy framework (utilizing ontology based data access) described in our work (cite). This implementation is a work in progress and it is missing several features to be used in production. However, all the functionality described in our work (cite) is implemented.
Policies are external instructions (very similar to event-condition action rules) that define a system's behaviours. However, they are "soft" rules that can be violated in case of a conflict. i.e. one policy may forbid opening a window if there is a cat in the room, while the other policy may oblige opening a window. In this case, one of the policies should be violated if there is no other way to ventilate the room. In the context of Internet of Things (IoT), policy frameworks can be used to develop digital assistants such as Siri, Alexa, or Google Home.
The software in the repository takes the following as inputs: a policy file that contains all the policies, an ontology that is used to describe those policies, and the database mappings of the ontology.
Then, the software can be used to detect conflicts between policies, to keep track of active policy instances, and to generate plans for executing obligations while minimizing policy violation costs. We note that converting action descriptions to PDDL is not in the scope of this work and PDDL actions should be provided to the planner. An overview of the architecture is provided below.
These instructions will let you run our project and examples on your local machine for development or testing purposes.
The project is being developed with Java 8.
The easiest way to run the project is installing Eclipse (or your favorite IDE) and importing the code as a Maven project. The IDE should automatically download all the dependencies.
We use the Fast Downward planner in our tests. We recommend using the source code provided in our repository, as we had to make a slight change (disabled verify_axiom_predicates function in /translate/normalize.py file) in the code. (This issue will be addressed in the future.) See the instructions to compile the planner. Be sure to compile the planner before you try running the smart home example.
The configuration file contains the following parameters that can be changed. All the paths written in this file should be relative paths. Below is the configuration file we use for our smart home example:
owlFile=resources/use_cases/smart_home/sspn-ql-rdf.owl
obdaFile=resources/use_cases/smart_home/sspn-ql-rdf.obda
obdaPropertiesFile=resources/use_cases/smart_home/sspn-ql-rdf.properties
ontologyIri=http://cs.ozyegin.edu.tr/muratsensoy/2015/03/sspn-ql#
policyFile=resources/use_cases/smart_home/policies.xml
plannerDomain=planner/fastdownward/domain.pddl
plannerProblem=planner/fastdownward/problem.pddl
plannerOutput=planner/fastdownward/sas_plan
plannerDir=planner/fastdownward/
plannerCommand=fast-downward.py domain.pddl problem.pddl --search "lazy_greedy([ff()], preferred=[ff()])"
smartHomeDB=resources/use_cases/smart_home/h2.sql
We provide two examples in this project; a smart home example that works with a single in-memory database and a smart mine example that works with data federation.