Integrating Adaptive Reference Vectors with Differential Evolution Algorithm for Multi-Objective Service Composition

Multi-objective service composition is increasingly crucial in distributed system engineering, especially in optimizing quality-of-service (QoS) attributes. This approach seeks to identify Pareto-optimal solutions regarding QoS attributes in distributed systems. As service requests increase, finding all Pareto-optimal solutions becomes excessively time-consuming. Although existing methods that approximate the Pareto-optimal set are scalable, they still require substantial enhancements in effectiveness. To overcome this challenge, this paper aims to refine the reference vectors to more uniformly guide the solutions towards the true Pareto-optimal set. Consequently, we propose a novel multi-objective differential evolution algorithm designed to efficiently identify solutions that demonstrate both proximity to and diversity within the Pareto-optimal set. This algorithm utilizes a reference vector-based cascade search to gather essential nondominated compositions efficiently. Additionally, it features an adaptive reference vector mechanism specifically designed to address the complexities of intricate multi-objective optimization problems. The robustness and effectiveness of our proposed algorithm have been rigorously validated using a real-world dataset, demonstrating its capability to improve service composition in distributed systems.