Smart Cities Research Center
Analyzing big data to create sustainable mobility systems.
A collaboration between UC Berkeley and Lawrence Berkeley National Laboratory.
The Smart Cities Research Center strives to sustainably and equitably improve mobility and quality of life in our cities. We do this through advanced quantitative modeling of urban systems, using that data to power the complex, interdisciplinary decision-making required to manage modern cities.
We focus on optimizing mobility, energy, productivity, regional economics, and quality of life in our cities by increasing mobility system efficiency, reducing cost, reducing fossil fuel use and increasing the effectiveness of transportation.
The pillars of our work are as follows:
Rich geospatial data
At the center of our work is Mobiliti, our cutting-edge software system that accurately simulates the movement of an entire population through a region’s road networks.
Unlike traditional simulation capabilities, Mobiliti is able to handle the incredible volume of data that comes with modeling millions of trips across an entire metropolitan system.
Just how well does Mobiliti scale? For the entire San Francisco Bay area (population ~8 million), the system can simulate a whole day’s worth of trips in under four minutes.
This game-changing performance opens up incredible possibilities.
We're experts in these areas
Sense-making of imperfect, massive geospatial datasets is foundational to everything else we do. We’re experts in efficiently determining staypoints, interpreting driver behavior, and even automatically detecting which intersections in a road network use stoplights.
The real power of Mobiliti is in being able to simulate emergent behavior when that equilibrium is disrupted. To do this effectively at the scale of billions of events per day, we’ve taken an innovative approach to parallel discrete event simulation – an algorithm through which the impact of sequential events on a system can be processed in parallel without losing the causality between those events.
The future of transportation management lies not only in understanding the state of the world, but also in proactively managing it. To this end, we’ve developed neural networks to accurately predict future highway traffic, and are investing in reinforcement learning approaches to manage dynamic traffic signal timing.
The GPS-based services we rely on today (Google Maps, Waze, Uber, Lyft) all optimize for minimizing vehicle travel time. But that approach comes with externalities – a famous example being quiet residential streets suddenly facing a deluge of Lyfts taking a shortcut. Our Mobiliti analytics allow us to explore the impacts of alternative routing strategies – on safety, on fuel consumption, and much more.