Generative Shopfloor Layout Design: Challenges and Proposed Modelling Approach

This paper explores the transformative impact of generative design principles on layout design and organization. The paper provides insights into the practical use of generative design in layout environments, emphasizing the optimization of spatial arrangements, workflow efficiency, and human-machine collaboration. The paper presents a novel generative approach proposed for optimizing manufacturing shop floor layouts, utilizing a three-stage process that integrates Variational Autoencoders (VAEs) and Generative Adversarial Networks (GANs). The first stage focuses on exploration and learning using VAE to generate a diverse pool of solutions, guided by a fitness profile encompassing characteristics such as travel distance, adjacency score, space utilization, ergonomic score, and aesthetics. The second stage employs GANs for intensive layout improvements, introducing a competitive training process where the generator aims to produce layouts surpassing the quality of previous iterations. The third stage fine-Tunes the best layouts identified by the GAN.A much more detailed model is required, and a VAE with higher resolution than the one used in the first phase. The manuscript presents a comprehensive methodology, paving the way for future research into dynamic layout adaptation in manufacturing settings.