Bi-objective service composition and optimal selection for cloud manufacturing with QoS and robustness criteria

As an important part of cloud manufacturing, service composition and optimal selection (SCOS) has attracted many scholars’ attentions. Although there are a large number of researches devoted to this field, plenty of challenges still remain, such as the difficulties of coping with real-life uncertainties and improving efficiency in making decisions. To this end, a bi-objective service composition and optimal selection (BoSCOS) problem which takes both QoS and robustness into account is firstly studied in this paper, its mathematical model is established after the two criteria are constructed. Then, a strengthened multi-objective gray wolf optimizer (SMOGWO) is developed to solve the above model with high efficiency, in which three improvement strategies are introduced to the original MOGWO: (1) the opposition based learning (OBL) strategy is adopted to improve the quality of initial population; (2) a leader update strategy is designed to explore better search leaders and balance the exploitation and exploration abilities; (3) an improved archiving strategy is developed to retain all possible dominant individuals, so as to improve the population diversity based on the Pareto dominance. Finally, the effectiveness of the proposed algorithm is verified on 8 BoSCOS problems with different scales and 17 benchmark functions, results show that the improvement strategies are effective and SMOGWO is superior to its competitors in terms of convergence and population diversity.