Knowledge Representation Methods for Smart Devices in Intelligent Buildings

   page       attach   
Giuseppe Loseto

Home and Building Automation (HBA) –in short known as domotics– is basically aimed at coordinating subsystems and appliances in inhabited spaces to provide increased levels of user comfort and manageability, to reduce energy consumption and minimize environmental impact. In latest years, the design of smart HBA environments is attracting efforts from several disciplines, including mobile and pervasive computing, wireless sensor networks, artificial intelligence and agent-based software, coalescing into a research area named Ambient Intelligence (AmI). AmI refers to a vision where people are surrounded by several kinds of objects and devices, able to manage and adapt themselves according to user profiles and environmental constraints in a seamless and unobtrusive way.
A crucial issue for feasible and effective AmI solutions lies in efficient resource/service discovery. Current HBA systems and standard technologies are still based on explicit user inputs over static operational scenarios, established during system design and installation. Consequently, they allow a low degree of autonomicity and flexibility. Several attempts have been made to overcome limits and constraints deriving from current features of HBA systems and applications. Studies and research basically target novel communication protocols and application features as a mean to by-pass the lack of autonomous capabilities in adapting to the user and environment requirements. In the recent research on this field, it has to be mentioned the exploitation of Artificial Intelligence (AI) techniques for such purposes. Particularly, several studies today attempt to by-pass restrictions in HBA through the adaptation and integration of Knowledge Representation formalisms and techniques originally conceived for the Semantic Web. Ontology languages, based on Description Logics, can be used to describe the application domain and available resources in a way that can support inference procedures and matchmaking processes, in order to satisfy users’ needs
and preferences to the best possible extent and adapt to context changes and evolutions.
This work aims to propose a solution for pervasive knowledge-based HBA systems which is grounded on the Semantic Web of Things (SWoT) framework. The SWoT is an emerging vision in Information and Communication Technology, integrating the Internet of Things and Semantic Web approaches. It exploits semantic annotations coming from a large numbers of heterogeneous micro-devices, each conveying a small amount of useful information, in order to embed intelligence into everyday objects and locations, so giving intelligence and expressiveness to devices and envinronments.
There, a general-purpose framework for HBA is proposed, supporting semantic-enhanced characterization of both user requirements and contextual resources. According to the pervasive computing paradigm, during ordinary activities the user should be able to simultaneously exploit information and functionalities provided by multiple objects deployed in her surroundings. To grant such autonomic processing capabilities and user transparency to the whole domotic environment, each device should autonomously expose its services and it should be also able to discover functionalities and request services from other devices.
Technologies and ideas have been borrowed from the Semantic Web initiative and adapted to HBA scenarios. Semantic Web languages provide the basic terminological infrastructure for domotic ubiquitous Knowledge Bases (u-KBs) which enable the needed information interchange. An enhancement to ISO/IEC 14543-3 EIB/KNX standard for building automation has been devised to this purpose. The integration of a semantic micro-layer within the KNX protocol stack enables novel resource discovery and decision support features, while preserving full backward compatibility. In this way, machine-understandable metadata characterize both home environment and user profiles and preferences, by means of annotations expressed in ontological formalisms based on Description Logics. The full exploitation of semantics in user needs and device/environment description has several benefits: (i) improved device interoperability; (ii) reasoning on descriptions to characterize environmental conditions (context) and to support advanced services through semantic-based matchmaking; (iii) improved scalability, flexibility and autonomicity with respect to current HBA standards.
To enable a user-transparent and device-driven interaction as opposed to common domotic solutions, in particular a multi-agent framework has been defined where requests coming from users and/or devices are collected by a home daemon acting as a negotiation mediator between users and home appliances.
Finally, to provide the needed pervasiveness to the system infrastructure, a Semantic Sensor Network (SSN) approach was therefore investigated, featuring: (i) a semantic-based backward-compatible extension of Constrained Application Protocol (CoAP) emerging IETF (Internet Engineering Task Force) standard; (ii) employment of non-standard inference services for resource discovery; (iii) adoption of World Wide Web Consortium (W3C) Semantic Sensor Network Incubator Group (SSN-XG) ontology to annotate data, events and device features.
The remainder of the work is organized as in what follows. Chapter 1 explains the SWoT vision describing the related architecture and the integration of pervasive computing technologies. Details about reasoning services and supported logic languages are also given. Chapter 2 describes the framework architecture and the proposed enhancements to KNX protocol. Moreover relevant related work is briefly surveyed along with a description of the implemented testbed and performance evaluation. In Chapter 3 the proposed multi-agent architecture is outlined with a case study about power management in HBA, while in Chapter 4 the SSN discovery framework is thoroughly described, also providing details about CoAP extensions and event mining. Finally, conclusion and future perspectives close the work.