In this paper we present results from the first use of neural networks
for real-time control of the high-temperature plasma in a tokamak fus
ion experiment. The tokamak is currently the principal experimental de
vice for research into the magnetic confinement approach to controlled
fusion. In an effort to improve the energy confinement properties of
the high-temperature plasma inside tokamaks, recent experiments have f
ocused on the use of noncircular cross-sectional plasma shapes. Howeve
r, the accurate generation of such plasmas represents a demanding prob
lem involving simultaneous control of several parameters on a time sca
le as short as a few tens of microseconds. Application of neural netwo
rks to this problem requires fast hardware, for which we have develope
d a fully parallel custom implementation of a multilayer perceptron, b
ased on a hybrid of digital and analogue techniques.