Real Time Semantic Interoperability in ad hoc Networks of Geospatial Databases : disaster Management Context
|Advisor:||Brodeur, Jean; Mostafavi, Mir Abolfazl|
|Abstract:||The recent technological advances regarding the gathering and the sharing of geospatial data have made available important volume of geospatial data to potential users. Geospatial databases often represent the same geographical features but from different perspectives, and therefore, they are semantically heterogeneous. In order to support geospatial data sharing and collaboration between users of geospatial databases to achieve common goals, semantic heterogeneities must be resolved and users must have a shared understanding of the data being exchanged. That is, semantic interoperability of geospatial data must be achieved. At this time, numerous semantic interoperability approaches exist. However, the recent arrival and growing popularity of ad hoc networks has made the semantic interoperability problem more complex. Ad hoc networks of geospatial databases are network that self-organize for punctual needs and that do not rely on any predetermined structure. “Traditional” semantic interoperability approaches that were designed for two sources or for a limited static number of known sources are not suitable for ad hoc networks, which are dynamic and composed of a large number of autonomous sources. Nevertheless, while a semantic interoperability approach designed for ad hoc network should be scalable, it is essential to consider, when describing semantics of data, the particularities, the different contexts and the thematic, spatial and temporal aspects of geospatial data. In this thesis, a new approach for real time semantic interoperability in ad hoc network of geospatial databases that address the requirements posed by both geospatial databases and ad hoc networks is proposed. The main contributions of this approach for real time semantic interoperability are related to the dynamic collaboration among user agents of different geospatial databases, knowledge representation and extraction, automatic semantic mapping and semantic similarity, and query propagation in ad hoc network based on multi-agent theory. The conceptual framework that sets the foundation of the approach is based on principles of communication between agents in social network. Following the conceptual framework, this thesis proposes a new model for representing coalitions of geospatial databases that aim at supporting the collaboration among user agents of different geospatial databases of the network, in a semantic interoperability context. Based on that model, a new approach for discovering relevant sources and coalitions mining based on network analysis techniques is proposed. Operators for the management of events affecting coalitions are defined to manage real times changes occurring in the ad hoc network. Once coalitions are established, data exchanges inside a coalition or between different coalitions are possible only if the representation of semantics of rich enough, and the semantic reconciliation is achieved between ontologies describing the different geospatial databases. To achieve this goal, in this thesis we have defined a new representation model for concepts, the Multi-View Augmented Concept (MVAC). The role of this model is to enrich concepts of ontologies with their various contexts, the semantics of their spatiotemporal properties, and the dependencies between their features. A method to generate MVAC concept was developed. This method includes a method for the extraction of the different views of a concept that correspond to the different contexts, and an augmentation method based on association rule mining to extract dependencies between features. Then, two complementary models to resolve semantic heterogeneity between MVAC concepts were developed. First, a gradual automated semantic mapping model, the G-MAP, discovers qualitative semantic relations between MVAC concepts using rule-based reasoning engines that integrate new matching criteria. The ability of this model to take as input a rich and complex representation of concepts constitutes the contribution of this model with respect to existing ones. Second, we have developed Sim-Net, a Description Logic- based semantic similarity model adapted to ad hoc networks. The combination of both models supports an optimal interpretation by the user of the meaning of relations between concepts of different geospatial databases, improving semantic interoperability. The last component is a multi-strategy query propagation approach for ad hoc network. Strategies are formalized with the Lightweight Coordination Calculus (LCC), which support interactions between agents based on social norms and constraints in a distributed system, and they represent the different strategies employed to communicate in social networks. An algorithm for the real time adaptation of strategies to changes affecting the network is proposed. The approach was implemented with a prototype using the Java JXTA platform that simulates dynamic interaction between peers and groups of peers. The advantages, the usefulness and the feasibility of the approach were demonstrated with a disaster management scenario. An additional contribution is made in this thesis with the development of the new notion of semantic interoperability quality, and a framework to assess semantic interoperability quality. This framework was used to validate the approach. This new concept of semantic interoperability quality opens many new research perspectives with respect to the quality of data exchanges in network and its impact on decision-making.|
|Document Type:||Thèse de doctorat|
|Open Access Date:||18 April 2018|
|Collection:||Thèses et mémoires|
All documents in CorpusUL are protected by Copyright Act of Canada.