The work carried out at Yale during the last three years has resulted in a number of neural network models for the identification and control of nonlinear dynamical systems. It has also raised numerous theoretical and practical questions concerning the prior information that will be needed to control the plant adaptively, as well as to assure its robustness in more complex environments. To address these question, it is proposed to investigate the following four related areas: (i) Extension of the methods developed to the control of nonlinear systems in the presence of bounded disturbances, and time- varying parameters; development of multilevel controllers for fault tolerant systems. (ii) New classes of plants for which stable adaptive algorithms can be generated. (iii) Properties of multilayer neural networks and radial basis function networks which are used as components in complex dynamical systems and their impact on overall performance. (iv) Theoretical basis for all the investigations carried out using results from nonlinear control theory.
