Modeling of urban form based on the 'Beady Ring' model (Hillier & Hanson, 1984) using deep reinforcement learning.

100 eps

BR_sample_vid03.mp4

500 eps

BR_sample_vid02.mp4

1000+ eps

BR_sample_vid.mp4

The cells could be in one of 3 different states at a time:

• An empty state (grey)
• A house cell (black)
• A street cell (white)

The agent gets a local observation in the form of an array with the states of the cells under the green area surrounding the current action cell (adjustable observation reach), or gets the full state of the environment as an image showing the states of all cells in a full observation version of the environment.