A parallel perspective for user-centered and QoS-aware service composition in the Internet of Things

Abstract

he rapid growth of the Internet of Things (IoT) has led to the proliferation of services, making Quality of Service (QoS)-aware service composition a critical challenge. This thesis addresses this issue through three complementary contributions. The first contribution presents a systematic literature review of QoS-aware service composition approaches, introducing a two-layered taxonomy that distinguishes between plan-based and autonomous approaches. This review identifies key limitations in the state-of-the-art, such as the lack of semantic matching consideration, the assumption of a prior existence of an abstract composition plan, limited scalability, and the use of fixed population sizes. These findings highlight the need for more effcient and adaptive approaches. To overcome these issues, the second contribution proposes the parallel differential evolution-based approach with population size reduction for QoS-aware services composition (PDE-QSC). By evolving two parallel sub-populations with distinct strategies and adaptively reducing population size, the PDE-QSC approach improves the composition quality and computation time compared to five baseline approaches. However, this approach still relies on the existence of an abstract plan and does not account for the semantic matching aspect. The third contribution addresses these remaining limitations by introducing two database concepts-driven approaches for autonomous QoS-aware semantic service composition (HCFDSSC and ESFDSSC). The proposed approaches leverage functional dependency theory to ensure semantic feasibility, reduce the search space, achieve fault tolerance in the case of service failures, and generate high-quality compositions. This thesis advances QoS-aware service composition in IoT by linking plan-based optimization and autonomous semantic approaches, opening new perspectives for scalable, adaptive, and resilient service systems.