Multiobjective cooperative multi-fitness in workflow scheduling problem

The optimisation of scientific workflows in cloud environments presents considerable challenges, primarily due to the inherent trade-offs between makespan and energy consumption. To address this, we propose Multi-Objective Cooperative Multi-Fitness (MOCMF), a novel mechanism that significantly enhances multi-objective evolutionary algorithms through a unique cooperative evaluation and recoding strategy. Diverging from existing multi-decoder approaches, MOCMF’s core innovation lies in its collaborative framework: heuristic decoders work in tandem to support a baseline decoding function, providing expert solutions that guide the Lamarckian recoding of chromosomes. Furthermore, MOCMF extends this cooperative evaluation to a multi-objective setting, where each heuristic decoder focuses on optimising a specific objective, leading to the generation of multiple distinct solutions per chromosome. Experimental results on data-intensive workflow benchmarks show that MOCMF improves the average Hypervolume by 32% and Inverted Generational Distance by 42% compared to a standard NSGA-II implementation, and by 7% and 6% respectively compared to its mono-objective cooperative variants. The proposed mechanism is also generalisable and potentially applicable to other multi-objective problems beyond workflow scheduling.