Optimal design of high-autonomy non-holonomic super neural networks

It is suggested that a self-development process of high-autonomy systems based on dynamic neural networks can be formulated within a framework of generalized nonholonomic dynamic systems and extended to reach the nature of dynamic interactions in complex design tasks involving some cognitive, intellectual invention and discovery or other decision processes. This feature is considered to be critical for any true autonomy; however it also introduces a potential for tremendous risk of unknown new events. To better determine some of the undesirable consequences, a concept of nonholonomic constraints for representing dynamic changes in relationships within neural networks and super neural networks is generalized to quantum nonholonomic constraints. This is intended to develop barrier mechanisms of a psychological nature in the mutual interactions of high-autonomy systems. A target-dedicated self-development of nonholonomic constraints is introduced. It is intended to provide mechanisms for optimal self-control development. In this formulation both supervised and unsupervised learning processes could be a part of the optimal self-control mechanisms