skip to main content


Caltech's prototype EV charging stations
Caltech's prototype EV charging stations include touch-screen LCDs to facilitate user interaction (left).
Credit: Credit: Caltech

A system that allows electric vehicles to replace their fossil-fuel counterparts and also helps stabilize power grid operation

Transportation and electricity production consume two-thirds of U.S. energy and account for almost 60 percent of greenhouse gases. But an electric vehicle (EV) can increase a household's electricity demand by 40 percent, which stresses the grid. And as California marches towards 100 percent clean electricity by 2045, wind and solar generation will challenge grid operation due to unpredictable weather conditions. These two problems are intertwined and must be tackled jointly.

A key element of the solution is smart, affordable, large-scale charging that will clean transportation, incorporate renewable energy, and facilitate stable grid operation.

Engineer and mathematician Steven Low has developed an adaptive sensor network that integrates real-time information about vehicle charge levels, available renewable energy, and grid conditions to optimize EV charging. Private enterprise has already commercialized this technology and captured a quarter of the large-scale charging market in the United States.

Steven Low, Frank J. Gilloon Professor of Computing and Mathematical Sciences and Electrical Engineering


Outdoor cameras take pictures automatically when they detect motion or body heat—day or night, rain or shine. Beery's algorithms can identify animals in these often dark or blurry photos and mark them with red boxes.

Systems that enable biologists to track animals without tagging, collaring, or even seeing them

Caltech computer science graduate student Sara Beery develops machine-learning algorithms that can pinpoint animals in images captured by unattended cameras. Her goal is to create a real-time model of global biodiversity that shows where species are and are not present.

Beery is part of a broad Caltech collaboration that focuses on methods of automated wildlife identification used by biologists and laypeople around the world. The researchers have created open-source AI models that detect animals in images, together with a system that learns to extract contextual clues—for example, about animals' behavior over time—from unlabeled images.

"The goal is for anyone to be able to put out a camera and to get real-time notifications of the animal species that are detected," Beery says. "This data helps scientists and citizen scientists and will improve global biodiversity models. We're able to leverage people's passions, their interests, their hobbies to directly provide research data."

Team Members
Sara Beery, graduate student, Computing and Mathematical Sciences
Elijah Cole, graduate student, Computing and Mathematical Sciences
Peter Kulits, undergraduate student
Pietro Perona, Allen E. Puckett Professor of Electrical Engineering


Shasta Lake, a California reservoir, in September 2014

A statewide model that enables communities to respond more effectively to increasing and more intense droughts

Reservoirs store rainwater for use during dry periods. But decisions about how much water to release are challenging. They must take into account the needs of farms, towns, hydropower plants, and fish and wildlife, in addition to complex factors such as weather patterns.

Sometimes, information is insufficient or arrives too late. This was the case when drought struck California from 2012 through 2015, for example, and some reservoir managers struggled to determine how much to reduce water releases, or even whether to discontinue them.

"There was a collective sense that responding to drought conditions on a per-reservoir basis was far from optimal and that an integrated statewide approach was needed," says Armeen Taeb (PhD '19), then a graduate student in electrical engineering.

With Caltech professor Venkat Chandrasekaran, JPL engineer Michael Turmon, and JPL hydrologist John Reager, Taeb created the first statewide empirical reservoir model. The model integrates economic statistics and weather and reservoir data with information about 55 major reservoirs to provide a dashboard of system-wide conditions.

Team Members
Venkat Chandrasekaran, Professor of Computing and Mathematical Sciences and Electrical Engineering
Michael Turmon, JPL Engineer
John Reager, JPL Research Scientist
Armeen Taeb (PhD '19)