Recent progress in the use of artificial neural networks (ANN's) as subsystems in control systems, in Japanese laboratories as well as by American researchers, has been based on the enhancement of these systems with system concepts from the brain, notably the cerebellum. This research project will assess the progress of Ann-based control system design, especially for robotic systems, which incorporates neurophysiological concepts and suggest new directions for future development. Improved applied control systems should have adaptive components to achieve high levels of sophistication and performance. Algorithms with neurological plausibility will be considered for implementation as ANN's. These will relate to intelligent neural network technology as it exists in the United States, Japan and elsewhere. It is expected that this assessment will identify the current directions of research as indicated in the literature. It is important to suggest new directions which are relatively undeveloped but offer considerable promise. Clearly, additional neuroscientific based subsystems can be identified from the literature. These subsystems and learning techniques can be developed and mathematically modeled for computer simulation to be studied for use in ANN control systems. These enhanced ANN designs will be applied to robotic control, in future work, in an effort to substantiate, and enhance the progress made to date (Miyamoto, Kawato, Setoyama & Suzuke, 1988).
