Smart cities: Inverse design of 3D urban procedural models with traffic and weather simulation

Ignacio Garcia Dorado, Purdue University

Abstract

Urbanization, the demographic transition from rural to urban, has changed how we envision and share the world. From just one-fourth of the population living in cities one hundred years ago, now more than half of the population does, and this ratio is expected to grow in the near future. Creating more sustainable, accessible, safe, and enjoyable cities has become an imperative. A city cannot longer be seen as a static set of buildings interconnected with roads. It is both a complex and interdependent dynamic system. Many disciplines, such as urban planning, traffic engineering, and architecture, have created approaches that try to design and model different aspects of a city facing challenging problems. However, due to its massive scale and complexity there has not been any attempt to develop a framework that addresses all these aspects at the same time. This is our challenge. We use an incremental approach to improve the realism of simulation and design: urban reconstruction, procedural generation, inverse procedural modeling, traffic engineering, and weather forecasting. We start with urban reconstruction that allows us to create detailed models for visualization and planning. Our methods focus on fast reconstruction and refinement of structures. Procedural modeling permits encapsulating the complex inter-dependencies within realistic urban spaces and enables users, who need not be aware of the internal details of the procedural model, to create quickly large complex 3D city models. Using machine-learning techniques, we can achieve real time interaction of high-level indicators. Vehicular traffic design has been carried out using aggregated simulations given that per-vehicle-simulation used to be too computationally expensive. That is, after observing that each car's behavior depends just on its current state and surrounding vehicles, we discretize each lane in the road network as a set of contiguous memory bytes. At each simulation step each car concurrently can check its surroundings to define its future state. Finally, we explore how weather is an essential aspect of a city and how we can use it to improve its design. We develop a fast but complete weather simulator that allow exploring interactively how certain city designs exert positive impacts on a city when weather changes. We also present additional work in generating 3D assets from photographs for such urban modeling environments. Initial applications of our work include creating enhanced and optimized 3D cities, simulating and visualizing urban space, traffic, and weather. Most of these applications have been developed in collaboration with academic, governmental, and industrial organizations. Our results include city models spanning up to 50 km square, traffic simulation over 300,000 simultaneous cars, and weather encompassing 2500 km square. We expect our efforts will ultimately increase the interdisciplinary collaboration in the development of better and smarter cities.

Degree

Ph.D.

Advisors

Aliaga, Purdue University.

Subject Area

Computer science

Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server
.

Share

COinS