Intelligent optimisation of traditional village element layout using generative adversarial networks

The conservation and optimisation of traditional villages has progressively been a major focus for intelligent applications with the fast development of artificial intelligence and information technology. Combining spatial multi-dimensional constraints to intelligibly optimise the layout of traditional villages through the adversarial training of generators and discriminators, this paper proposes a traditional village element layout optimisation method GAN-PLS based on generative adversarial network (GAN) model. Introduced into the GAN-PLS model, the gradient penalty technique helps to increase the stability of the training process and optimisation effect. Particularly outperforming conventional optimisation techniques in terms of convergence speed, generation effect and stability, the GAN-PLS model shows good performance in spatial layout optimisation, cultural element retention and ecological conservation by means of comparison studies. At last, this work addresses the shortcomings of the model and suggests future directions like dataset expansion, computational efficiency enhancement, and multi-dimensional constraints addition.