Backpropagation and neurocontrol: a review and prospectus

The author reviews highlights of the emerging field of neurocontrol and describes how backpropagation could be used in a new generation of controllers suitable for problems involving thousands or millions of control variables and a high degree of noise and nonlinearity. A summary is presented of the five dominant paradigms or architectures now used in this field. The adaptive critic paradigm and how it fits in with backpropagation and dynamic feedback are discussed in more detail. Four adaptive critic architectures are outlined.<<ETX>>