ABSTRACT
Neurocontrol is a subset of the larger field of control theory, which designs systems for a broad spectrum of applications, ranging from simple regulators (like thermostats or muscle neurons) to optimal decision-making in complex environments (as in the brain as a whole system). Neurocontrol, like classical control and artificial intelligence, includes general designs for three basic types of task: cloning, tracking and optimization. Neural cloning systems copy the input-output behavior of human experts or automatic controllers. Tracking systems may be regulators, or systems to make a robot arm follow (track) a desired path in space, etc. Optimization over time may be used to solve tracking problems, with improved stability, or to solve planning problems which require real intelligence. This section compares the practical advantages and disadvantages of a wide variety of control designs, neural and otherwise, ranging from simple regulators through to designs which begin to provide an explanation of intelligence in brain circuits.
