Sign up to save your podcasts
Or
Share Revolutionizing Nuclear Fusion with Magnetic Control and Reinforcement Learning with Dr. Felici
Share to email
Share to Facebook
Share to X
Share to email
Share to Facebook
Share to X
Or
Revolutionizing Nuclear Fusion with Magnetic Control and Reinforcement Learning with Dr. Felici
In this insightful episode, we're delighted to welcome Dr. Felici, an innovator pushing the boundaries of sustainable energy production through nuclear fusion. Dr. Felici introduces us to the complex world of magnetic confinement in tokamak configurations, highlighting the promise and challenges this path toward sustainable energy presents.
One of the core challenges in this field is shaping and maintaining high-temperature plasma within the tokamak vessel. Dr. Felici discusses the necessity of high-dimensional, high-frequency, closed-loop control using magnetic actuator coils and the complications posed by diverse requirements across a wide range of plasma configurations.
Delving into his groundbreaking work, Dr. Felici introduces an autonomous learning architecture for tokamak magnetic controller design. This approach elegantly meets control objectives specified at a high level while satisfying physical and operational constraints, offering unparalleled flexibility and generality in problem specification. What's even more exciting is how this approach dramatically reduces the design effort required to produce new plasma configurations.
Dr. Felici takes us on a journey of how this new approach was applied to produce and control a diverse set of plasma configurations on the Tokamak à Configuration Variable (TCV). The approach enabled the creation of advanced configurations, including elongated, conventional shapes, as well as negative triangularity and 'snowflake' configurations. The episode also delves into how the team demonstrated sustained 'droplets' on TCV, where two separate plasmas are maintained simultaneously within the vessel.
By showcasing the potential of reinforcement learning in accelerating research in the fusion domain, this conversation with Dr. Felici sheds light on the most challenging real-world systems where reinforcement learning has been successfully applied. If you're interested in sustainable energy, advanced learning systems, or the intersection of the two, this is an episode you won't want to miss!
https://doi.org/10.1038/s41586-021-04301-9
View all episodes
By Catarina CunhaIn this insightful episode, we're delighted to welcome Dr. Felici, an innovator pushing the boundaries of sustainable energy production through nuclear fusion. Dr. Felici introduces us to the complex world of magnetic confinement in tokamak configurations, highlighting the promise and challenges this path toward sustainable energy presents.
One of the core challenges in this field is shaping and maintaining high-temperature plasma within the tokamak vessel. Dr. Felici discusses the necessity of high-dimensional, high-frequency, closed-loop control using magnetic actuator coils and the complications posed by diverse requirements across a wide range of plasma configurations.
Delving into his groundbreaking work, Dr. Felici introduces an autonomous learning architecture for tokamak magnetic controller design. This approach elegantly meets control objectives specified at a high level while satisfying physical and operational constraints, offering unparalleled flexibility and generality in problem specification. What's even more exciting is how this approach dramatically reduces the design effort required to produce new plasma configurations.
Dr. Felici takes us on a journey of how this new approach was applied to produce and control a diverse set of plasma configurations on the Tokamak à Configuration Variable (TCV). The approach enabled the creation of advanced configurations, including elongated, conventional shapes, as well as negative triangularity and 'snowflake' configurations. The episode also delves into how the team demonstrated sustained 'droplets' on TCV, where two separate plasmas are maintained simultaneously within the vessel.
By showcasing the potential of reinforcement learning in accelerating research in the fusion domain, this conversation with Dr. Felici sheds light on the most challenging real-world systems where reinforcement learning has been successfully applied. If you're interested in sustainable energy, advanced learning systems, or the intersection of the two, this is an episode you won't want to miss!
https://doi.org/10.1038/s41586-021-04301-9