User-adapted service selection in pervasive computing environments: Taking quality of service and context awareness into consideration

The seamlessly integration of heterogeneous devices embedded with pervasive services provides indeed a higher degree of ubiquity to users, though this benefit may turn into a drawback if users feel overloaded when they are confront with too many choices with similar functionality. Addressing to this inevitable problem, some researchers concentrate on the service side and present ontology-based description models which enrich QoS aspect semantics. They also propose common selection strategies that decide on the best services by evaluating solely quality related properties. Concerning their approaches, this paper argues that services with the best qualities are not necessarily best suitable services according to users' needs. We believe that additionally users' local contextual restrictions and subjective preferences on services bound devices' properties must be considered in order to achieve the best suitable service selection result. Motivated by the challenge of user-adapted suitable service selection in pervasive computing environments, we take use of the semantic web technology and aim at developing a comprehensive selection framework. In this paper, first, we present our designed comprehensive knowledge base in the framework which is composed of two ontology-based models: one enhances additional semantics on service nonfunctional properties (QoS and Context of resources) for description; another supports user-side either accurate or ambiguous criteria expressions. Second, conforming to our suitable selection strategy, we present a selection algorithm in the framework which multi-evaluates user-side criteria against potential service properties and makes suggestions for best suitable services. At last, we propose an agent based architecture to integrate our framework as a whole. In three-stage experiments, we test our framework separately with 15 participants. We evaluate users' conceptual understanding of our two models. Then, we compare how well our selection mechanism succeeds in selecting best-suitable services comparing to selection with generic or solely QoS-supported strategies. And not least, we validate the selection performance under two given difficult selection situations using Paired T-Test. After analyzing the results of our experiments, we finally conclude our research work and conducts the future work