NSF Award
9704067
This Small Business Innovation Research Phase II project will develop a unique adaptive system identification (SI) method that will enable reconfigurable modeling of a complex nonlinear dynamic system. This unique SI method will apply to a wide range of operating regimes as well as adapt to slow changes in the system being modeled. The SI method will be demonstrated in an adaptive critic applied to a difficult real-world problem. It will use an Ontogenetic (variable-structure learning) neural network to learn to compensate for the difference between a linear and nonlinear system model, producing results within a given accuracy of the system's actual performance. Hosted on an Multiple Instruction Multiple Data (MIMD) neural network processor, it will be computationally very fast. The SI method will be used in adaptive critic neurocontrol of an experimental aircraft testbed. Commercial applications of the SI technology under development in this Phase II program include process control, financial forecasting, and aircraft flight control.
