Research Professor, Robotics Institute Carnegie Mellon University
Jeff Schneider is researching how to use machine learning to control fusion reactions.
CMU News online
2025-07-11
From reimagining AI data centers to modernizing and securing the electric grid, CMU researchers are working on practical solutions to pressing challenges in how the U.S. produces, moves and secures energy.
“Many of the world's grand challenges like a sufficient food supply, clean water availability and climate issues are actually just energy problems," says Jeff Schneider, a research professor in CMU’s School of Computer Science. "Nuclear fusion and its promise of limitless clean energy would solve many of them."
CMU News online
2025-07-11
Carnegie Mellon University experts are developing practical solutions for a fast-changing energy system.
"This moment in energy history is unique because AI will play a central role in accelerating the development, optimization and operation of our energy sources,"says Jeff Schneider, Research Professor in Carnegie Mellon University's Robotics Institute. "Generative AI models now usefully represent and exploit the domain knowledge for these tasks while reinforcement learning and other discovery and control learning algorithms can now solve real-world problems much faster than the old human-only method of making technological progress.
Carnegie Mellon University online
2025-03-24
Using AI could help unlock a new potential source of energy to solve that problem, including work by Jeff Schneider, research professor in the School of Computer Science, and his research team studying nuclear fusion.
1 min
Carnegie Mellon University experts are developing practical solutions for a fast-changing energy system. Their work modernizes infrastructure, accelerates innovation and harnesses AI for a more efficient and resilient future at a moment when the stakes for national competitiveness and public well-being have never been higher. Learn what CMU experts have to say about their Work That Matters.
Ph.D.
Computer Science
B.S.
Computer Science
2023 | US Coast Guard AI Boot Camp Online
2023 | General Electric EDGE Symposium Niskayuna, NY
2023 | CMS Large Hadron Collider Annual Meeting Pittsburgh, PA
US11953901
An autonomous vehicle includes one or more sensors for detecting an object in an environment surrounding the autonomous vehicle and a vehicle computing system comprising one or more processors receiving canonical route data associated with at least one canonical route, and controlling travel of the autonomous vehicle based on sensor data from the one or more sensors and the canonical route data associated with the at least one canonical route. The at least one canonical route comprises at least one roadway connected with another roadway in a plurality of roadways in a geographic location that satisfies at least one route optimization function derived based on trip data associated with one or more traversals of the plurality of roadways in a geographic location by one or more autonomous vehicles.
US11835951
2021-10-25 Assigned to UBER TECHNOLOGIES, INC. reassignment UBER TECHNOLOGIES, INC. ASSIGNMENT OF ASSIGNORS INTEREST (SEE DOCUMENT FOR DETAILS). Assignors: RADOSAVLJEVIC, VLADAN, HUANG, TZU-KUO, CHOU, FANG-CHIEH, CUI, Henggang, DJURIC, Nemanja, LIN, TSUNG-HAN, NGUYEN, Thi Duong, SCHNEIDER, Jeff
US11635764
Systems, methods, tangible non-transitory computer-readable media, and devices associated with the motion prediction and operation of a device including a vehicle are provided. For example, a vehicle computing system can access state data including information associated with locations and characteristics of objects over a plurality of time intervals. Trajectories of the objects at subsequent time intervals following the plurality of time intervals can be determined based on the state data and a machine-learned tracking and kinematics model. The trajectories of the objects can include predicted locations of the objects at subsequent time intervals that follow the plurality of time intervals. Further, the predicted locations of the objects can be based on physical constraints of the objects.
2024
Developing and maintaining safe operating regimes are necessary for tokamak-based commercial fusion power plants while tearing modes pose a threat to steady state operation. Future machines plan to have active tearing mode (TM) control by driving current using electron cyclotron heating (ECH), but plans based on previous experiments have the restrictions of wastefully using ECH for TM suppression or needing to wait …
2024
Typically the rampdown phase of a shot consists of a decrease in current and injected power and optionally a change in shape, but there is considerable flexibility in the rate, sequencing, and duration of these changes. On the next generation of tokamaks it is essential that this is done safely as the device could be damaged by the stored thermal and electromagnetic energy present in the plasma. This works presents a procedure for automatically choosing experimental rampdown designs to rapidly converge to an effective rampdown trajectory. This procedure uses probabilistic machine learning methods paired with acquisition functions taken from Bayesian optimization. In a set of 2022 experiments at DIII-D, the rampdown designs produced by our method maintained plasma control down to substantially lower current and energy levels than are typically observed.
2023
Deep reinforcement learning (RL) has shown immense potential for learning to control systems through data alone. However, one challenge deep RL faces is that the full state of the system is often not observable. When this is the case, the policy needs to leverage the history of observations to infer the current state. At the same time, differences between the training and testing environments makes it critical for the policy not to overfit to the sequence of observations it sees at training time. As such, there is an important balancing act between having the history encoder be flexible enough to extract relevant information, yet be robust to changes in the environment. To strike this balance, we look to the PID controller for inspiration. We assert the PID controller's success shows that only summing and differencing are needed to accumulate information over time for many control tasks.
2022
Many potential applications of reinforcement learning (RL) are stymied by the large numbers of samples required to learn an effective policy. This is especially true when applying RL to real-world control tasks, eg in the sciences or robotics, where executing a policy in the environment is costly. In popular RL algorithms, agents typically explore either by adding stochasticity to a reward-maximizing policy or by attempting to gather maximal information about environment dynamics without taking the given task into account. In this work, we develop a method that allows us to plan for exploration while taking both the task and the current knowledge about the dynamics into account. The key insight to our approach is to plan an action sequence that maximizes the expected information gain about the optimal trajectory for the task at hand.
2022
Alfvén eigenmodes (AEs) are an important and complex class of plasma dynamics commonly observed in tokamaks and other plasma devices. In this work, we manually labeled a small database of 26 discharges from the DIII-D tokamak in order to train simple neural-network-based models for classifying AEs. The models provide spatiotemporally local identification of four types of AEs by using an array of 40 electron cyclotron emission (ECE) signals as inputs. Despite the minimal dataset, this strategy performs well at spatiotemporally localized classification of AEs, indicating future opportunities for more sophisticated models and incorporation into real-time control strategies.
