Date of Award
Santa Clara : Santa Clara University, 2021.
Computer Science and Engineering
Artificial intelligence and machine learning have recently become a hot topic in terms of software solutions to complex problems. Every year, new prototypes and projects are created to solve specific problems, ranging from self-driving vehicles to facial recognition. One project which interested our team was OpenAI Five, which created an artificial intelligence agent to play the complex online competitive game DotA 2. We wanted to create our own agent to design optimized city layouts in the game Cities: Skylines. By doing so, we hope to illustrate the viability of using artificial intelligence as a tool in urban planning for real cities.
To create the artificial intelligence agent, we took advantage of the fact that we were using a video game as an environment for the agent. We were able to customize and control the environment that the agent was operating in, and allow the agent to reliably read data on the game state and perform actions to change the game state. With these prerequisites, the agent was able to begin the process of reinforcement learning to progressively find more optimal urban design solutions for the game environment.
Duncan, Carter; Cunningham, Jack; Wang, Andrew; and Kennedy, Alexander, "Urban Planning Optimization via “Cities: Skylines”" (2021). Computer Science and Engineering Senior Theses. 215.