CONTEXT-POLICY-CONFIGURATION

Paradigm of Intelligent Autonomous System Creation

Oleksiy Khriyenko, Sergiy Nikitin

Industrial Ontologies Group, Agora Center, University of Jyväskylä, P.O. Box 35(Agora), FIN-40014 Jyväskylä, Finland

Vagan Terziyan

Industrial Ontologies Group, MIT Department, University of Jyväskylä, P.O. Box 35(Agora), FIN-40014 Jyväskylä, Finland

Keywords: Policy, Policy-based Configuration, Autonomous System Collaboration, Context Awareness of System

Behaviour.

Abstract: Next generation of integration systems will utilize different methods and techniques to achieve the vision of

ubiquitous knowledge: Semantic Web and Web Services, Agent Technologies and Mobility. Nowadays,

unlimited interoperability and collaboration are the important things for industry, business, education and

research, health and wellness, and other areas of people life. All the parties in a collaboration process have

to share data as well as information about actions they are performing. During the last couple of years,

policies have gained attention both in research and industry. Policies are considered as an appropriate means

for controlling the behaviour of complex systems and are used in different domains for automating system

administration tasks (configuration, security, or Quality of Service (QoS) monitoring and assurance). The

paper presents Semantic Web driven approach for context-aware policy-based system configuration.

Proposed Context-Policy-Configuration approach for creation of intelligent autonomous systems allows

system behaviour modification without source code change or requiring information about the dependencies

of the components being governed. The system can continuously be adjusted to externally imposed

constraints by policy determination and change.

1 INTRODUCTION

During the last tens year’s humankind utilizes

computers and numerous research results in the area

of information technologies to build intelligent

systems that help people in various domains:

professional activities, entertainments, social sphere

and etc. Experts build such systems in different

domains to solve various tasks. But, all of them use

models of data and knowledge representation as a

basis for system creation.

With an intensive development of the Internet

and very fast growing amount of information and

data world wide, semantic technologies have

become very popular. To achieve the vision of

ubiquitous knowledge, the next generation of

integration systems will utilize different methods

and techniques. These include Semantic Web

(Semantic Web, 2001, Berners-Lee et al., 2001) and

Web Services, Agent Technologies, Mobility

(Curbera et al., 2002, Clabby, 2002), and

WebServices (Ankolekar et al., 2002, Paolucci et.

al., 2002, FIPA, 2001). Semantic technologies are

viewed today as a key technology to resolve the

problems of interoperability and integration within

the heterogeneous world of ubiquitously

interconnected objects and systems. But still, aspects

such as context and proactivity of these resources

and systems are quite in demand nowadays and

should be considered more comprehensively.

Semantic technologies are a qualitatively stronger

approach to interoperability than contemporary

standards based approaches.

At the same time, to make system really

intelligent, dynamic and autonomous, we have to

utilize Agent Technologies. The vision of autonomic

computing emphasizes that the run-time self-

manageability of a complex system requires its

components to be, to a certain degree autonomous

themselves. Software agent technologies will play an

198

Khriyenko O., Nikitin S. and Terziyan V. (2010).

CONTEXT-POLICY-CONFIGURATION - Paradigm of Intelligent Autonomous System Creation.

In Proceedings of the 12th International Conference on Enterprise Information Systems - Artiﬁcial Intelligence and Decision Support Systems, pages

198-205

 SciTePress

important part in building such complex systems.

Agent based approach to software engineering is

also considered to be facilitating the design of

complex systems. When it comes to developing

complex, distributed software based systems, the

agent based approach is advocated (Jennings, 2001).

From the implementation point of view, agents are

the next step in the evolution of software

engineering approaches and programming

languages, the step following the trend towards

increasing degrees of localization and encapsulation

in the basic building blocks of the programming

models (Jennings, 2000).

To be smart, system should be able to take into

account current state of environment, react on

changes of it and behave accordingly. Thus, there is

a need to elaborate such a framework of system

development, which allow us to build an influence

model of certain situation not only on single entity,

but also on a system as whole.

From other side, to provide an ability to the

system to be automatically controllable, policy based

approach may be highly valuable. Policies are “rules

governing choices in the behaviour of a system”

(Damianou et. al., 2001). They contain the logic for

guiding decisions during the execution of the

system. Through policy, people can precisely

express bounds on autonomous behaviour

(permissions) and expectations of performance

(obligations) in a way that is consistent with their

appraisal of an agent’s competence in a given

context. Policies allow changing the behaviour of a

system without changing low-level code, creating

more adaptable systems whose behaviour can be

altered dynamically. The ability to change policies

dynamically means that poorly performing agents

can be immediately brought into compliance with

corrective measures.

Policies can be used in different areas and

domains. For example, policies have been used for

access control, network and quality-of-service

management, user preferences, operational policies,

storage management, system configuration, self-

management, multi-agent systems, etc. The policy

community has been researching topics like policy

specification, management, enforcement, reasoning,

negotiation, refinement, discovery and many others.

Policies can be expressed at different levels, referred

to as a policy hierarchy, ranging from high-level

abstract policies over specification-level policies to

low-level configuration policies.

Among other benefits of policy-based

approaches, there are: reusability, efficiency,

extensibility, context-sensitivity, verifiability,

support for both simple and sophisticated

components, and external reasoning about

component behavior. Policies can be used to

explicitly express agreements, conventions,

precedents, and salience conditions that help make

automated components more effective players. They

can be used to enforce bounds and expectations that

increase interpredictability, they can be used to

establish and maintain common ground, and their

ability to be imposed and adjusted at runtime

enables dynamic directability.

2 CONTEXT-POLICY

CONFIGURATION PARADIGM

Developing and maintaining large-scale, distributed

applications is a complex task. Middleware has

traditionally been used to simplify application

development by hiding low-level details and by

offering generic services that can be reused and

configured by application developers. However,

middleware technology has not kept up with the

growing demands that emerge in the digital society:

the scale of distributed applications is rapidly

increasing, the range of users that compose and

configure applications has expanded significantly,

the increased scope of distributed applications has

also resulted in more advanced application

composition scenarios.

We are basing our research on UBIWARE

Platform. The UBIWARE Platform is a development

framework for creating multi-agent systems. It is

built on the top of the Java Agent Development

Framework JADE

, which is a Java implementation

of IEEE FIPA specifications. The name of the

platform comes from the name of the research

project, in which it was developed. In the

UBIWARE project

, a multi-agent system was seen,

first of all, as a middleware providing

interoperability of heterogeneous (industrial)

resources and making them proactive and in a way

smart.

2.1 Proactive Goal-driven Resource as

a Main Entity of Any System…

A resource of a new Web is a proactive goal-driven

dynamic entity that adequately and proactively

JADE - http://jade.tilab.com/

UBIWARE project - http://www.cs.jyu.fi/ai/OntoGroup/

UBIWARE_details.htm

CONTEXT-POLICY-CONFIGURATION - Paradigm of Intelligent Autonomous System Creation

199

Picture 1: Agent environment structure.

reacts on changes within its external environment or

within itself. As a consequence of resource

dynamism and proactivnes, environment itself

becomes more dynamic. Following GUN (Kaykova

et. al., 2005) vision (Global Understanding

Environment, where all the resources of the virtual

and the real world are connected and interoperate

with each other) that considers “everything as a

resource” (even abstract entities) with correspondent

semantic annotation and adapter, accompanied with

an agent to be proactive, dynamic and autonomous.

Accordingly to semantic extension of agent

programming language Semantic Agent

Programming Language (S-APL) (Katasonov and

Terziyan, 2007) that solves such issues as

description of rules and beliefs (representing the

knowledge needed for playing the role) and

understanding of the semantics of the rules, the

meaning of predicates used in those rules by all the

parties involved while using first-order logic as the

basis for an APL, we are considering three agent

language constructs: Belief, Goal and Behavior.

Figure (Figure 1) shows us a structure of an

agent environment. Belief Storage is represented by

statements about environment, resource and agent

(as well as about other resources and agents). All

this information can be considered as preconditions

and input data for rules (behaviours) execution.

Then Rule Engine all the time is checking

availability of the rules that can be performed and

result changes within the beliefs as well as actions

with correspondent beliefs changes. Resource’s

behaviour results to changes in the Resource itself

and to changes in the environment. In return,

changes in the environment influence on the

Resources. Thus, such dynamism and proactivnes

bring context-awareness to the system, and more and

more statements and behaviours become context-

dependent. Thus, in this architecture the trigger that

runs agent behaviour is a goal that should be

specified for the agent. Based on specified goal, a

behaviour planning process builds the behaviour

model that is presented by appropriate set of

behavioural rules that leads agent towards the set

goal. Later such an abstract plan should be bound

with particular instances from the agent beliefs or

those that can be obtained via inference in runtime.

Such rule based behavior representation present us

rule based programming approach where conditional

part of the rule is written in a way of semantic

description that brings much more flexibility

comparing to hardcoded programming.

Applying this approach we are able to build a system

with two different levels of

programming/administration. First level is the level

of “advanced user” programming/administration and

implies building of the rules to reach different goals

that cover particular domain. It means that we define

certain domain by Ontology of Goals - set of

possible abstract goals that can be reached by

ICEIS 2010 - 12th International Conference on Enterprise Information Systems

200

we do not program system in a hardcoded way, but

build it able to change internal functionality and

behaviour on the fly when context is changed.

2.2 Role-based Policy Control

of a System

Generally we deal with a system with big amount of

entities (Resources) with own behaviours and goals. To

be able to control the system on general level, we have

to put constraints/policies on separate entities as well as

on the system in whole. System that is based on full

unlimited knowledge (Ontology and Rules) and has an

ability to make any actions is a Global (Mother)

System. This system is unrestricted by any specific

goal and able to behave towards achieving all possible

goals. Any organization, union, company, society,

group, individual and etc. can be considered as a sub

system that plays certain Role, which restricts it with

particular set of goals and knowledge/resources used

for goal achievement (see Figure 3). Here Role of a

system is a context that configures it via applying

correspondent policy. Thus, any context is a kind of

Role that implies (re-)configuration of a system.

Creation of Roles is a duty of “advanced

user/administrator” that creates correspondent set of

policies and rules of their application. In case, Role

has been applied and there is still a need to make

goal definition more concrete, then “high-level

user/administrator” can be asked to do this.

Coming back to the approach of autonomous goal-

driven Resource, application of new Role to the

Resource Agent means creation of new working area

for it, configured accordingly to correspondent

Policy that restricts its’ degree of freedom for

planning and execution processes. In this case

Resource can play several Roles and share available

resources based on own configured knowledge that

corresponds to the Role-related area. Concerning the

intelligent part of the Resource, the main aspect is

decision making. It is an ability to decide which one

among possible behaviours/actions is the best one in

particular situation, is profit estimation of one or

another plan among set of possible to be performed.

To make a decision we have to weight different

actions and deduct which of them are more

profitable/advantageous for the Resource depending

on its’ goal. Such technique is widely used in Game

Theory

that attempts to mathematically capture

behavior in strategic situations, in which an

individual's success in making choices depends on

the choices of others. There is a possibility to build

Game Theory

http://en.wikipedia.org/wiki/Game_Theory

different models to rank behaviours (from simple to

very complex calculation methods). These methods

should take into account context of current situation

to simplify calculation process and provide us more

relevant results. In another words, methods should

follow applied



-type policy (vector of priorities

of possible behaviours) defined by Role. It may

happen that Roles become contradicted in sense of

contradicted policies. In such situation Agent should

refuse one or several contradicted Roles to avoid any

contradictions in the system, or play all Roles if

contradictions are not strong and concern only

priorities of behaviours and goals. In the last case we

have a deal with nested Policy-based control and

new type of policy that regulates priority between

lower level Policies.

2.3 Policy Models in Use

Now, when we highlighted several areas and tasks

that need context-dependent policy-based control, let

us come to elaboration of a general vision of

Context-Policy-Configuration paradigm and show

models in use.

In the last decade, several policy description

languages have been developed, mostly designed for

specific purposes, including:

 Ponder (Damianou et. al., 2001) is a policy

language for specifying authorization and

obligation policies in the context of

distributed networks and systems;

 The eXtensible Access Control Markup

Language (XACML) (OASIS, 2005)

standard includes an XML-based policy

language for specifying authorization

policies;

 KAoS (Uszok et. al., 2003) and Rei (Kagal

et. al., 2003) are semantic policy languages

for expressing authorization and obligation

policies. They both allow the inclusion of

external ontologies, defining the semantics

of domain-specific concepts;

 The Web Service Policy framework (WS-

Policy) (IBM et. al., 2006) provides an

extensible grammar for expressing non-

functional requirements for interacting with

a web service;

 GlueQoS (Wohlstadter et. al., 2004) is an

extension of WS-Policy for specifying

quality-of-service (QoS) features of web

services.

Expectations regarding the new generation of Web

depend on the success of Semantic Web technology.

Resource Description Framework (RDF) is a basis

for explicit and machine-readable representation of

semantics.

ICEIS 2010 - 12th International Conference on Enterprise Information Systems

202

Figure 3: Role-based System creation.

Policy System Ontology

RestrictionContainer

D&R_RestrictionContainer

W_RestrictionContainer

basisSystem_is

hasRestriction

hasConfiguredOntology

rdfs:domain

rdfs:range

rdfs:domain

rdfs:range

rdfs:domain

rdfs:range

rdfs:subClassOf

Container

Property

Class Resource

rdfs:subClassOf

rdfs:type

rdfs:subClassOf

rdfs:type

rdfs:Statement

…

rdfs:Statement

…

RDFS

CPC

extension

applied to

rdfs:range

rdfs:domain

Figure 4: Initial part of Context-Policy-Configuration extension of RDFS.

To be compatible with widely used technology

we extend RDF Schema with some classes and

properties for policy description. Figure 4 shows us

initial part of CPC-extension of RDFS. In the

platform we utilize N3 representation in S-APL

language.

Let us follow an example to show main

principles of policy based system configuration. For

example (see Figure 5) consider a System –

“GreenFactory” as a subsystem of System -

“Factory” with only difference that “GreenFactory”

utilized only green kind of energy: Nuclear-, Hydro-,

Wind-, Sun-energy, etc. Here we can use a D&R-

type policy that restricts the ontology of mother-

system (“Factory”) and redefines a range of the

“useEnergy” property for the “GreenFactory”.

Figure shows us the range of the “useEnergy”

property in “Ontology #i” that is used by system -

“Factory #n”. The range is presented by list of

different kinds of energies (Oil-, Wood-, Coal-,

Nuclear-, Hydro-, Wind-, Sun-energy). Statement #1

states that “Policy #n” is applied to the system

“GreenFactory #m” in case if this system plays the

role “Role #k”.

CONTEXT-POLICY-CONFIGURATION - Paradigm of Intelligent Autonomous System Creation

203

Statement #1

subject

objectpredicate

Policy #n

applied to

GreenFactory #m

inContext

ContextContainer #1

Statement #2

subject

object

predicate

GreenFactory #m

hasRole

Role #k

Statement #3

subject

object

predicate

Factory #k

hasConfiguredOntology

Ontology #i

rdfs:Statement

rdfs:range

property: useEnergy

Oil-energy

Wood-energy

Coal-energy

Nuclear-energy

Hydro-energy

Wind-energy

Sun-energy

basisSystem_is

Factory #k

hasRestriction

D&R_RestrictionContainer #1

hasRestriction

W_RestrictionContainer #1

Policy #n

rdfs:Statement

rdfs:range

Nuclear-energy

Hydro-energy

Wind-energy

Sun-energy

property: useEnergy

rdfs:Statement

buyEnergyFrom

Supplier #j

System

rdfs:Statement

hasRelevance

0.7

Policies are considered as an appropriate means for

controlling the behaviour of complex systems.

They are used in different domains for

automating system administration tasks, such as

configuration, security, or Quality of Service (QoS)

monitoring and assurance. Context-Policy-

Configuration approach for creation of intelligent

autonomous systems is allowing modifying system

behaviour without changing source code or requiring

information about the dependencies of the

components being governed. The system can

continuously be adjusted to externally imposed

constraints by changing the determining the policies.

This research presents a policy-based approach

for supporting the high-level configuration of

systems, integrated into the middleware platform.

Policies are high-level, declarative statements

governing choices in the behaviour of a system. Our

“policy-driven middleware” extends the traditional

middleware architecture with an extra layer that

hides complexity when possible and enables

simplified application development and maintenance

by offering the means to express, validate and

enforce policies.

ACKNOWLEDGEMENTS

This research has been performed as part of

UBIWARE project in Agora Center (University of

Jyvaskyla, Finland) funded by TEKES and industrial

consortium of Metso Automation, ABB, Fingrid,

Inno-W and Hansa Ecuras. We are very grateful to

the members of “Industrial Ontologies Group” for

fruitful cooperation.

REFERENCES

Ankolekar, A., Burstein, M., Hobbs, J.R., Lassila, O.,

Martin, D. L., McDermott, D., McIlraith, S.A.,

Narayanan, S., Paolucci, M., Payne, T.R. and Sycara,

K., (2002). DAML-S: Web Service Description for the

Semantic Web, URL: http://www-2.cs.cmu.edu/

~terryp/Pubs/ ISWC2002-DAMLS.pdf.

Berners-Lee, T., Hendler, J. and Lassila, O., (2001). The

Semantic Web. Scientific American 284(5), 34-43.

Clabby, J., (2002). Web Services Executive Summary,

URL: http://www-106.ibm.com/ developerworks/

webservices/library/ws-gotcha/?dwzone= webservices.

Curbera, F., Dufler, M., Khalaf, R., Nagy, W., Mukhi, N.,

Weerawarana, S., (2002). Unraveling the Web

Services Web: An introduction to SOAP, WSDL and

UDDI, Internet computing.

Damianou, N., Dulay, N., Lupu, E. and Sloman, M.,

(2001). The ponder policy specification language.

Proceedings of the 2nd International Workshop on

Policies for Distributed Systems and Networks, 2001.

FIPA, (2001). FIPA Interaction Protocol Library

Specification Specification, FIPA00025. URL:

http://www.fipa.org/specs/fipa00025/

Jennings, N., (2000). On agent-based software

engineering. Artificial Intelligence 117(2), 277–296

Jennings, N., (2001). An agent-based approach for

building complex software systems. Communications

of the ACM 44, 4 (2001) 35–41.

IBM, BEA Systems, Microsoft, SAP AG, Sonic Software,

and VeriSign, (2006). Web services policy framework

(WS-Policy), March 2006.

Kagal, L., Finin, T. and Joshi, A., (2003). A policy

language for a pervasive computing environment. In

Proceedings of the 4th International Workshop on

Policies for Distributed Systems and Networks, 2003.

Katasonov, A. and Terziyan, V. (2007). SmartResource

Platform and Semantic Agent Programming Language

(S-APL), In: P. Petta et al. (Eds.), Proceedings of the

5-th German Conference on Multi-Agent System

Technologies (MATES’07), 24-26 September, 2007,

Leipzig, Germany, Springer, LNAI 4687 pp. 25-36.

Kaykova, O., Khriyenko, O., Kovtun, D., Naumenko, A.,

Terziyan, V. and Zharko, A., (2005). General

Adaption Framework: Enabling Interoperability for

Industrial Web Resources, In: International Journal on

Semantic Web and Information Systems, Idea Group,

ISSN: 1552-6283, Vol. 1, No. 3, July-September 2005,

pp.31-63.

Khriyenko, O. and Terziyan, V., (2006). A Framework for

Context-Sensitive Metadata Description. In:

International Journal of Metadata, Semantics and

Ontologies, Inderscience Publishers, ISSN 1744-2621,

Vol. 1, No. 2, pp. 154-164.

OASIS, (2005). eXtensible Access Control Markup

Language (XACML) version 2.0, OASIS standard,

February 2005.

Paolucci, M., Kawamura, T., Payne, T. R., Sycara, K.,

(2002). Importing the Semantic Web in UDDI. URL:

http://www-2.cs.cmu.edu/~softagents/papers

/Essw.pdf

Semantic Web, (2001). URL: http://www.w3.org/2001/sw/

Uszok, A., Bradshaw, J., Jeffers, R., Suri, N., Hayes, P.,

Breedy, M., Bunch, L., Johnson, M., Kulkarni, S. and

Lott, J., (2003). KAoS policy and domain services:

Toward a description-logic approach to policy

representation, deconfliction, and enforcement. In

Proceedings of the 4th International Workshop on

Policies for Distributed Systems and Networks, 2003.

Wohlstadter, E., Tai, S., Mikalsen, T., Rouvellou, I., and

Devanbu, P., (2004). GlueQoS: middleware to

sweeten quality-of-service policy interactions. In

Proceedings of the 26th International Conference on

Software Engineering, 2004.

CONTEXT-POLICY-CONFIGURATION - Paradigm of Intelligent Autonomous System Creation

205