CONNECT: Active Behavioral Interfaces for Seamless Network Integration

The CONNECT project is an ongoing research project 2006 - 2009, funded by the Research Council of Norway through the strategic programme VERDIKT. The project is a collaboration between the Interventional Centre at Rikshospitalet and the PMA group at the Department of Informatics. The main objective of the project is to develop a framework for the modeling and analysis of heterogenous networks in which devices may connect seamlessly in a semantically meaningful way. The framework is successful if it is able to address correctness issues for seamless adaptation of heterogenous networks, that are relevant and useful in practice. This will be assessed through a case study of biomedical sensor networks for patient monitoring.

 

Populærvitenskaplig sammendrag (in Norwegian)

CONNECT er et samarbeid mellom Institutt for informatikk ved Universitetet i Oslo og Intervensjonssenteret ved Rikshospitalet. CONNECT ser på grunnleggende utfordringer for sømløs integrasjon av programvareløsninger i heterogene nett ved å utvikle en meta-infrastruktur for kunnskap om omgivelse som vi kaller Active Behavioral Interfaces. Disse grensesnittene lar programvarekomponenter som mangler kunnskap om sin omgivelse, fungere sømløst i ulike omgivelser. Grensesnittene vil utføre viktige oppgaver for interaksjon som filtrering og konvertering av innkommende data, såvel som tilpasning av utgående data slik at både mengden av og kvaliteten på utgående data passer med omgivelsen.

Meta-infrastrukturen skiller design av komponenter fra design av grensesnitt og støtter høynivå beskrivelse av nettverk, samt monitorering av nettverkstraffikk. Dermed får grensesnittene informasjon om nettverket som tillater standardiserte løsninger for grensesnittene. I CONNECT fokuserer vi på modellerings- og designnivået og bygger et rammeverk for høynivå programmering og analyse av komponenter, gransesnitt og nettverk ved hjelp av verktøy for spesifikasjon, simulering, prototyping, modellsjekking og analyse. Komponenter og grensesnitt blir beskrevet i det objekt-orienterte språket Creol, mens nettverk modelleres i koordineringsspråket Reo. Utprøving av språk og verktøy skjer ved en omfattende case studie om pasient monitorering i heterogene nett, ledet av Intervensjonssenteret.

 

Short project overview

The project addresses basic challenges in seamless integration of software solutions for heterogeneous net, by proposing a meta-infrastructure for environment awareness based on a novel notion of Active Behavioral Interfaces. This notion of active behavioral interfaces will enable software components without environmental awareness to function seamlessly in different environments. More specifically, the active behavioral interface of a component will perform filtering, adaptation and conversion of incoming data, allowing the component to understand inputs intended for other kinds of components, and furthermore, perform adjustment and adaptation of outgoing data, allowing the quantity and quality of outgoing messages and message streams to fit the environment. In particular, the capacity and bandwidth of net connections are taken into consideration.

A meta-infrastructure will separate the design of active behavioral interfaces from that of the components themselves, and include a rich language for describing networks of different kinds. The meta-infrastructure will involve monitoring of the network traffic, and thereby provide information about the net to the component interfaces, allowing standardized interface solutions. The project will focus on the modeling and design level, building a framework for high level programming and analysis of components, interfaces and networks. The concepts will be explored, formalized, and implemented by tools oriented towards executable specifications, simulation, prototyping, model checking, and analysis. Components and interfaces are described in an object-oriented manner, designing a program structure useful for production level software solutions.

Approach and methodology

In order to address the problem of semantically meaningful composition of embedded devices in heterogeneous networks, we consider the integration of two different but complementary formal models of reconfigurable distributed systems:

In order to study end-to-end seamlessness of heterogeneous networks in this project, we start by integrating Creol and Reo, extending Creol objects with broadcast facilities. The two approaches are complementary: While Creol objects make no assumptions about the network, Reo networks make no assumptions about the components of the network. Consequently, Reo connectors may be used to adjust and combine the communication paths in a heterogeneous network in order to accommodate services required by Creol objects. This integration will be based on the notion of abstract behavioral interface, to be developed in this project. A behavioral interface will define
  1. the basic functionality (abstract syntax),
  2. sets of adaptive transformations, and
  3. sets of different protocols for communication.
Notions of abstract behavioral interfaces exist both for Creol and for Reo. The challenge is to combine these notions in a way which allows automated composition and seamless adaptation in order to allow devices to connect to a heterogeneous network in a semantically as well as syntactically meaningful way.

Formal approach and tool support

The research conducted in this project will have a formal basis, but we are also interested in the practical applicability of the work. Formal methods need tool support. The modeling language will be based on a formally defined and executable semantics. The semantics will be used to build a tool which supports rapid prototyping and validation of ad hoc networks, focusing on analyzing the effect of reconfigurations in the ad hoc network. The usefulness of the language and its possibilities for dynamic adaptability will be tested against a case study addressing a heterogeneous network for monitoring patients based on biomedical sensors, containing wireless as well as wired subnets. The case study will also be used to extend the tool with specialized simulation strategies which specifically highlights aspects of seamless adaptability, thereby providing valuable feed-back on the relevance of modeling concepts and tools.

We propose to build a prototyping and validation tool by representing the semantics of the modeling language in rewriting logic and execute specifications on the Maude platform. We have good experience in using rewriting logic and Maude as a prototyping and validation platform. An executable semantics for Creol objects, which supports runtime object update, has been defined in rewriting logic, and several case studies have been developed. It remains to integrate Reo and a representation of active behavioral interfaces into the operational semantics.

Case study: Heterogeneous nets for patient monitoring

The framework for the modeling and analysis of heterogenous networks developed in the CONNECT project will be applied to biomedical sensor networks in a collaboration between the Department of Informatics at the University of Oslo and the Interventional Centre at Rikshospitalet.

It is desirable that patient monitoring systems can operate seamlessly in different networking environments and can connect to different devices and tools. Several applications require that a combination of multiple biomedical sensors measuring physiological parameters in different locations on the tissue. Other applications may require that a number of different biomedical sensors measuring different physiological parameters. There are several aspects that can be considered for effective deployment of such a system: energy consumption, number of sensor nodes and days of use, distance-dependent path loss and shadowing, information routing, process estimation quality, node density, transmission protocol. In clinical applications, the location of sensor nodes and the environment are often defined. Context awareness, i.e., some knowledge of the environment, will provide information to maximize overall utility of the sensor network. Realistic models need to take considerations such as interferences from other biomedical devices as well as home appliances, path-loss due to signal transmission through tissues, and low power emission and radiation.

 

Collaborators:

University of Oslo (PMA group)

Interventional Centre, Rikshospitalet

Centrum voor Wiskunde en Informatica (CWI)

Einar Broch Johnsen Ilangko Balasingham Farhad Arbab
Olaf Owe   Frank de Boer

PhD students

Joakim Bjørk   
Fatemeh Kazemeyni

 




Funded by

< ifi | databehandling| pma >


Last modified 01.10.2007.