| Deriving from pervasive computing concept, Ambient Intelligence(AmI) fusescyberspace and physical space to provide computing services anytime and anywherein a nonintrusive way, (which means) that users are assisted proactively by a digitalenvironment surrounding them. Furthermore, AmI emphasizes on intelligentinteraction and personalization. Personalization of AmI transforms the traditionalstandardized service model,"one size fits all", into a proactive, adaptive anduser-oriented service model, solves the problem of"overload services", satisfies theuser need to a maximum extent, enhances the user experience, and provides keytechnical support for building the future information society.Although personalization has been researched massively and applied in the fieldsof searching engine, e-commerce, digital libraries, etc., those existing techonologiesare difficult to adapt to AmI enviorment in which user need is different from person toperson, context varies dynamically, and services are mostly heterogeneous. Therefor,it is extremely complicated to construct a service system to meet personalized userneed in high dimension AmI-Space composed of user, context and service. At present,the research of AmI personalization is still in its infancy, both at home and abroad,asking for a general architecture as well as related technologies to guide theimplementation of AmI-PS.According to state-of-the-art of AmI and personalized servicing in other fields,this dissertation mainlyfocuses on the following research contents:(1) Based on abstract of common function features, the research on service modeand systemic architecture achieves the unified understanding of AmI-PS to lead thedevelopment of AmI-PS;(2) Study on domain knowledge and personalized knowledge representation inAmI. User's individual characteristics can be formally described to form the basis forpersonalized servicing;(3) Study on reasoning techniques to find the optimal services for meeting userneeds best;(4) Study on learning techniques for adapting dynamic changes of AmI such asusers, context, services, to provide services timely and accurately; (5) Location information is very critical for AmI-PS. Due to the lack ofpositioning techniques adapting to AmI, study of a novel positioning approaches hasan important role in AmI-PS.This dissertation first researches comprehensively on recommendationtechnology, context awareness technology, knowledge representation, web service,mobile agent, wireless location, etc. as related supporting technologies of AmI-PS tounderstand and design the system better. Then several key issues of AmI-PS aresolved through above aspects of deeply researching. The main contributions andinnovations are as follows:(1) Simulating user's natural personalized servicing process, a novel distributedservice model is proposed to divide AmI-PS process into three stages: needsreasoning, needs meet reasoning, personalized servicing parameter reasoning. It haslots of virtues as following: fully showing personalization, high usage of distributedcomputing resource, mobility, robustness, extensibility, privacy protection. Based onabstracting and partitioning common system functionality, a general systemframework and hierarchy structure are proposed to describe the system's overalloperating mechanism and work processes of internal module and reduce thecomplexity of the system implementation, laying a foundation of other keytechniques.(2) Top ontology and domain ontology are implemented in OWL to represent theshare knowledge of AmI-PS. need ontology is used to associate users with services.AmI-HPM personalized approach is proposed to model personalized information bymany factors such as personalized need model, individual need satisfaction model andpersonalized servicing parameters model.(3) Through fusing shortest-path algorithm and degree of need satisfaction, apersonalized servicing planning theory is proposed to realize personalized reasoningwith the abilityof uncertain reasoning and deterministic reasoning. Firstly it forms thegraph of services by service demand ontology and service dependence. Then optimalservice path is found through computing the length of service path by compositiondegree of need-matching and QOS of service.(4) On combination of pairwise comparison and ANN, PC-ANN personalizedmodel weight learning algorithm is proposed. Leveraging user intuitive andconvenient method of pairs-wise analysis to rapidly build model weight, it not only overcome ANN's long supervised training process, but also quickly adapts to changesof user's characteristics according to user feedback. Furthermore, some classic singlefactors evaluation functions are raised including user habits, preferences and interests.(5) Multi-hops collaboration positioning algorithm (AmI-MHCL) and locationoptimization algorithm based on Least Squares error analysis (LSEAO) are proposed.AmI-MHCL uses multi-hops anchors to collaborate positioning, increasingpositioning ratio in low-density-anchors situation, especially for nodes at networkedge. Utilizing least squares error analysis, LSEAO could compute position with errorrange to reduce the impact of error propagation significantly and increase positioningaccuracy.(6) AmI-H is designed and implemented as AmI-PS prototype system in familyscenario. Firstly, AmI-H provides representative temperature control and personalizedTV recommendation service. Applying above proposed theory and approaches it fullyreflect the characteristics of the user. The rationality and correctness of theory andapproaches proposed are validated, in addition it provides a reference model for otherAmI-PS development.In this dissertation, key techniques of AmI from overall framework to detailapproaches are researched to enable AmI-PS meeting user personalized needs andenhancing user's experience in an accurate and efficient way. It adapts the traits ofAmI in which users, context, services all are massive, changing, and diverse. Theresults of research have great theoretical and practical significance to the developmentof future information society.
|
PDF Full Text Request