Back to top

Large-Scale Data Analytics

The past decade saw the explosion of data, with the convergence of sensing, computing, and communication on single platforms, and the transportation landscape is just beginning to see the profound changes to come from the use of this data by new technologies in the fields of automation, energy, planning, and operations. A new era is beginning in which the transportation sector has the opportunity to enable decision making based on large-scale data analytics. The Institute’s track record of contributions in the field of data and analytics includes freeway monitoring, crowdsourced traffic data, and simulations for urban planning. Today, data is the dominant culture at ITS. But in order to usher this field into its next generation, developments are needed beyond transportation, particularly in machine learning, vision, cloud computing, control, and optimization — all areas with strong expertise at UC Berkeley.

ITS has already built on numerous campus initiatives and institutes at the forefront of data analytics to bring the benefits of these advances to transportation. Partners include the Simons Institute, a new theoretical computer science institute exploring unsolved problems at the limits of computation, and the Algorithms Machine People (AMP) Lab, which works at the intersection of  machine learning, cloud computing, and crowdsourcing. ITS is also collaborating with Systems and the SmartCities academic programs, both hosted by the Department of Civil and Environmental Engineering. New faculty in these programs integrate machine learning — especially deep learning — communication, control, behavioral modeling, and other disciplines, shifting the demographics at ITS. Through the Large-Scale Data Analytics growth area, ITS will use these extraordinary assets to advance data-driven transportation solutions.

Alexei Pozdnukhov (left) works on his SmartBay Project with a student. The project, which aims to be a decision-support tool, applies data mining and machine-learning techniques to cell phone data and social network signals  to analyze and model Bay Area activities and travel habits.

Q&A Alexei Pozdnukhov: Smart Cities
Assistant Professor, Civil and Environmental Engineering; Co-Director, Smart Cities Research Center

Transportation interests: I work on developing new scalable methods of data analysis that will be used for planning and operation of transportation in future cities. To increase efficiency and reduce the environmental footprint of transportation, new tools and methods are required to leverage the insights hidden in data streams.

Background: I started my career as a machine-learning researcher working on pattern recognition, primarily in computer vision. However, modern cities are increasingly dense and data-rich, penetrated with IT systems and smart infrastructure that generate media streams. You start seeing patterns in urban dynamics that are rich and detailed. It inspired me to switch my focus on what we call urban data analytics.

Pressing research questions: One of the challenges for me is to understand how these heterogeneous data sources can be leveraged to close the control loops over city infrastructure. It is not sufficient to be able to just observe changes. We should be able to manage everyday operations as well as the longer-term evolution of the complex urban systems that we have engineered. It will soon become challenging to operate urban infrastructures as they become increasingly dependent on data flows and the algorithms that process the data. The resilience of this data-centric design will need to be addressed, especially as it brings in human behavior, with all its inherent uncertainties as more and more data come directly from handheld and mobile devices. Finally, infrastructure systems that were previously decoupled are becoming more interconnected and interdependent, introducing an extra layer of complexity. For example, the electric grid is becoming coupled with transport via a growing fleet of electric vehicles.