Methodology for Self-Adaptively Solving Multi-Objective Scheduling Problems

Scheduling practices are critical decision-making processes that substantially influence the overall performance of cloud and manufacturing environments. Therefore, scheduling problems have been a primary concern of practitioners and scholars in this field for decades. The majority of scheduling problems are known NP-hard optimization problems. Hence, heuristic and improvement methods have been conventionally adopted to address scheduling concerns. Heuristic methods exhibit a light execution time but fail to sustain high solution quality for solving complex problems. Improvement methods deliver high-quality solutions but are associated with high computational effort. Therefore, a scheduling methodology is presented that efficiently facilitates the combined utilization of heuristic, metaheuristic, and deep reinforcement learning methods to solve scheduling problems in cloud and manufacturing environments. Since most industrial scheduling problems are subject to multi-objective optimization measures, the methodology addresses scheduling concerns considering system efficiency and customer satisfaction objective measures. Parallelization and scalability technologies have been adopted to design and develop the presented artifact to achieve computational efficiency.