A large-scale data-oriented intelligent system for urban growth simulation

Abstract

With more than 50% of world population living in cities (as per UN report), urbanization is one of the pressing problems, which has been recognized by the United Nations. One of the outcomes of large-scale urbanization in cities is uncontrolled urban growth which has many potentially adverse effects on the environment. Due to this, various studies have been conducted on methods of simulating urban growth which assists in making appropriate decisions and plans. However, the studies involving urban growth simulation are very much dependent on human ingenuity because factors behind urban growth and their relationships are varied and complex. Therefore, scholars generally do not have consensus on these aspects which leads to difficulty in designing large-scale systems. Furthermore, the ability of machines to extract patterns from data has improved significantly in the past few decades with the advent of deep neural networks and high-performance computing architectures. In this work, we propose an intelligent system to model urban growth by minimizing the human ingenuity constraint from the simulation process. The system is based on machine learning, deep learning methods, and high-performance computing systems. Presently, the system uses the satellite remote sensing imagery and transportation network data for the urban growth simulation. We have experimented our proposed architecture on data gathered for the city of Mumbai, India, and have found encouraging performance compared to the existing learning-based methods.