Policy manifold generation for multi-task multi-objective optimization of energy flexible machining systems

Contemporary organizations recognize the importance of lean and green production to realize ecological economic benefits. Compared with existing optimization methods, multi-task multi-objective reinforcement learning (MT-MORL) offers an attractive means address dynamic, multi-target process-optimization problems associated Energy-Flexible Machining (EFM). Despite recent advances in learning, realization accurate Pareto frontier representation remains a major challenge. This article presents generative manifold-based policy-search method approximate continuously distributed for EFM optimization. To this end, multi-pass operations are formulated as part multi-policy Markov decision process, wherein machining configurations witness dynamic changes. However, traditional Gaussian distribution cannot accurately fit complex upper-level policies. Thus, multi-layered generator was designed map high-dimensional policy manifold from simple without performing calculations. Additionally, hybrid training approach is proposed handle mode collapse large task difference observed during improvement generalization performance. Extensive computational testing comparisons against baseline methods have been performed demonstrate improved quality efficiency algorithm.