Magnetic control of tokamak plasmas through deep reinforcement learning.
Jonas DegraveFederico FeliciJonas BuchliMichael NeunertBrendan D. TraceyFrancesco CarpaneseTimo EwaldsRoland HafnerAbbas AbdolmalekiDiego de Las CasasCraig DonnerLeslie FritzCristian GalpertiAndrea HuberJames KeelingMaria TsimpoukelliJackie KayAntoine MerleJean-Marc MoretSeb NouryFederico PesamoscaDavid PfauOlivier SauterCristian SommarivaStefano CodaBasil DuvalAmbrogio FasoliPushmeet KohliKoray KavukcuogluDemis HassabisMartin A. RiedmillerPublished in: Nat. (2022)
Keyphrases
- reinforcement learning
- optimal control
- control problems
- magnetic field
- control policy
- control strategies
- machine learning
- multi agent
- markov decision processes
- feedback control
- adaptive control
- robot control
- control theory
- learning process
- learning algorithm
- control strategy
- control policies
- function approximation
- real time
- data acquisition
- dynamic programming
- control system
- neural network