Since heavily non-linear and/or very complex processes stillpose a problem for automatic control, they can often behandled easily by human operators. The book describes results from ten years of research on learning control loopswhich imitate these abilities. After discussing the differencesto adaptive control some background on human information processing and behaviour is put forward and some learning control loop structure related to these ideas is shown.The ability to learn is due to memories, which are able tointerpolate for multi-dimensional input spaces between scattered output values. A neuronally and mathematically inspired memory lay out-are compared and it is shown that theylearn much faster thanbackpropagation neural networkswhich can also be used. For the learning control loop different architectures are given. Their usefulness is demonstrated by simulation and results from applications to real pilot plants. The book should be of interest for control engineers as well as researchers in neural net applicationsand/or artificial intelligence. The usual mathematical background of engineers is sufficient.

Springer Book Archives

Basic considerations.- Microintelligence.- Macrointelligence.