|Tipo di tesi||Tesi di dottorato di ricerca|
|Titolo||Approcci basati su agenti e semantica per lo sviluppo di infrastrutture per scenari intelligenti|
|Titolo in inglese||Agent- and semantic-based approaches to develop infrastructures for intelligent scenarios|
|Settore scientifico disciplinare||ING-INF/05 - SISTEMI DI ELABORAZIONE DELLE INFORMAZIONI|
|Corso di studi||Scuola di D.R. in INFORMATION AND COMMUNICATION TECHNOLOGIES (ICT)|
|Data inizio appello||2012-02-27|
|Disponibilità||Accessibile via web (tutti i file della tesi sono accessibili)|
L'attività di ricerca presentata in questo lavoro copre due diversi argomenti: approcci basati sulla semantica e sugli agenti per sviluppare infrastrutture per scenari intelligenti. Il primo argomento riguarda i motori di ricerca basati sulle parole-chiave e il secondo sistemi multi-agente per la gestione delle emergenze territoriale.
The research activity presented in this work covers two different topics: semantic- and agent-based approaches to develop infrastructures for intelligent scenarios. The first topic covers the keyword-based search engines and the second, multi-agent systems for disaster management. Keyword queries over relational databases offer a convenient alternative to traditional SQL, especially in case of large, complex or unknown schemas. The challenge in answering a keyword query over a relational database is to find in what database structures the items described by the keywords are modelled, and how these structures relate to each other. Current approaches apply information retrieval technique to collections of data previously gathered and indexed. This technique limits the use of these search engines in many interesting application scenarios, including information integration systems, in which the instance data is not directly available for building the required indexes. For this reason, the research activity was addressed to develop a new approach exploiting only meta-data information of the data sources, and using a set of auxiliary information to convert keyword queries into semantically meaningful SQL expressions. The meta-information are extracted from the DBMS and enriched by an human operator or exploiting by ontologies such as WordNet. The novelty of this approach lays on the way each keyword is mapped to a schema element (such as table and attribute) or to a value (attribute domain). The core of this technique is an extended version of the Hungarian algorithm which significantly reduces the number of possible keyword interpretations that need to be considered. For each interpretation, the set of possible path connections are discovered using graph techniques. For each path the SQL query string is build, executed and data retrieved and sent to the user. To perform tests on this new approach, a graphical user interface prototype has been implemented. Finally, a set of experimental evaluations illustrates its efficiency and effectiveness. The other topic faced during the research activity is the coordination and the cooperation in territorial management scenarios using the multi-agent technology. Territorial emergencies and large-scale disasters are one of the most critic activities in healthcare systems. In such scenarios, a very quick response is necessary in order to reduce the serious damage of people health involved in disasters and to safe as many lives as possible. The criticality of the situation needs a very fast response, coordination, cooperation and decision making. Many are the actors in these scenarios such as medical staff, firemen and policemen. All these actors must cooperate and coordinate in order to reach their target. Decision making must be fast and accurate. To face such requirement the multi-agent technology has been employed and in particular the JADE platform. The architecture designed, called Ubimedic2 (an extension of Ubimedic), provide a P2P system managed by agents that autonomously exchange the necessary information and organize the rescue operation. This approach is novel and has not been previously faces in the disaster management scenarios. Agents share information using FIPA compliant messages and organize the operations based on the BDI (Belief-Desire-Intention) model. Each component of the architecture has been implemented with the main functionality it must provide. Based on the architectural design, the framework has been developed in Java and a GUI prototype is used to simulate real case studies. Moreover, two applications for smart-phones have been developed to perform tests on real devices and evaluate the feasibility of the approach. Finally the integration with the PIM technology has been done in order to improve the coordination process among agents, reducing the number of messages exchanged and exploiting the strong mobility to compact the coordination logic in a single process.