Automated urban planning aware spatial hierarchies and human instructions

Traditional urban planning demands experts to spend much time producing an optimal plan under many architectural constraints. The remarkable imaginative ability of deep generative learning provides hope for renovating this domain. Existing works are constrained by: (1) neglecting human requirements; (2) omitting spatial hierarchies, and (3) lacking samples. We propose a novel, human-instructed planner fill these gaps implement two practical frameworks. In the preliminary version, we formulate task into encoder–decoder paradigm. encoder is learn information distribution surrounding contexts, instructions, land-use configuration. decoder reconstruct configuration associated functional zones. Although it has achieved good results, generation performance still unstable due complex optimization directions decoder. Thus, cascading adversarial network (GAN) in paper, inspired by workflow experts. first GAN build zones based on instructions contexts. second will produce considering built Finally, conducted extensive experiments case studies validate effectiveness superiority our work.