By combining concepts from game theory and estimation, Stanford researchers created LUCIDGames, a computational technique that can predict and plan adaptive trajectories for autonomous vehicles. LUCIDGames, an inverse optimal control algorithm is able to estimate the other agents' objective functions in real time, and incorporate those estimates online into a receding-horizon game-theoretic planner.
Lanthanum nickel oxide or LNO, belongs to a promising class of catalytic materials known as perovskites, used to split water into hydrogen and oxygen is the first step in generating hydrogen fuel. Several manufacturers have produced electric cars powered by hydrogen fuel cells. A research team from Stanford and the DoE's SLAC National Accelerator were able to see how surface transformation boosts the catalyst activity and could potentially apply to other materials with the ability to fine tune their behavior.
A TomKat Center for Sustainable Energy Innovation Transfer grant recipient, Aeromutable, is looking to streamline the trucking fleet efficiency to build a healthier future and more sustainable economy. Their aerodynamic tool uses air injection--what they call "virtual surfaces" to alter the aerodynamics flow along the truck. These air fins can be adjusted constantly using an AI-run controller which calibrate the tail to best suit the weather and road conditions in real time.
This interactive map developed by the American Energy Society and Sally Benson, Precourt Family Professor along with a former student of Benson's, Scott Jespersen, is a useful tool to innovators for the energy transition.
Stanford researchers have developed a new and principled approach to calculating the value of data used to train AI models based on a Nobel Prize-winning economics method that improves determining the value of individual data points or datasets. It can also help AI systems designers identify low value data that should be excluded as well as high value data worth acquiring and even used to reduce bias in AI systems.
Stanford scientists have designed an AI-capable chip by harnessing eight hybrid chips, each with its own data processor built next to its own memory storage. Called the Illusion System, the hybrids are tricked into thinking they're one chip. In simulations, systems with 64 hybrid chips were shown to run AI applications seven times faster than current processors, using one-seventh as much energy.
An untapped source of stored energy, EV batteries could power homes when sections of the grid are shut down to prevent wildfires. This could help residents weather power shutoffs and lessen pushback against grid shutdowns. EVs could also support community microgrids--local self-sufficient electricity supply networks. This article looks at steps that would need to be taken to overcome barriers before California could tap cars and charging stations for energy.
Anthony Corso, Ph.D. candidate presents his thesis defense on safety validation approaches. Corso is advised by Prof. Mykel Kochenderfer, Stanford Intelligent Systems Lab. Slides are available here.
To search for new materials, conventional methods employed by material scientists involve incrementally improving what is already known. Using a new method combining AI and high-throughput experimental techniques, researchers from the National Institute for Standards and Technology, the University of Maryland and the U.S. Department of Energy's SLAC National Accelerator Lab have developed a way to search a wider range of possibilities more efficiently speeding the discovery of desired new materials.
Forbes' 2021 "30 Under 30" includes three current Stanford students along with five recent alumni developing energy and sustainability-related technologies. Alumna, Annie Baldwin, seen here at left, MS MBA '19 was recognized for her work as director of strategy at eIQ Mobility which helps companies with vehicle fleets analyze driving patterns to determine which ICE vehicles should be replaced with EVs
