Abstract:
This thesis addresses the problem of QoS-aware service composition in large-scale CyberPhysical-Social Systems (CPSS) and the Internet of Things (IoT). These environments are highly dynamic and interconnected, which makes ensuring optimal Quality of Service (QoS), balanced energy
consumption, and adaptation to user mobility particularly challenging. Existing approaches often fail to
account for these dimensions simultaneously, resulting in reduced service availability and lower composition quality. To overcome these limitations, the first contribution introduces the Group Teaching-based
Energy-efficient and QoS-aware Composition Algorithm (GT-EQCA) for IoT environments. By formulating the problem as a multi-objective combinatorial optimization and selecting only the top-k most
relevant services, the GT-EQCA algorithm reduces the computation time while maintaining a high QoS
level. It employs the Group Teaching Optimization (GTO) algorithm, which avoids the need for hard
parameter tuning and ensures scalability. The experimental results show improvements of up to 76%
in composition time, 88% in energy efficiency, and 28% in QoS utility. However, the mobility aspect
is not addressed in this algorithm. To deal with this limitation, the second contribution proposes the
Learning-based Swarm optimization-aware Service Composition Algorithm (LS-SCA) by jointly considering mobility, energy, and QoS during the composition process. Using a realistic mobility model and
a two-phase learning-based swarm optimizer, the LS-SCA algorithm reduces computation time while
improving composition quality. The results show 23% higher QoS utility, 28% less energy consumption,
and 40% higher service availability compared to the literature baselines. This thesis proposes adaptive,
energy-efficient, and mobility-aware algorithms for scalable and reliable service composition in dynamic
IoT and CPSS environments.
