The research seeks a way to spontaneously update urban design outcomes in response to constant changing design input, using state of art machine learning algorithms - a multi-agents based reinforcement learning system.