Approach to adaptive neural net-based H∞ control design

An approach is investigated for the adaptive neural net-based H, control design of a class of nonlinear uncertain systems. In the proposed framework, two multi-layer feedforward neural networks are constructed as an alternative to approximate the nonlinear system. The neural networks are piecewisely interpolated to generate a linear differential inclusion model by which a linear state feedback H, control law can be applied. An adaptive weight adjustment mechanism for the multi-layer feedforward neural networks is developed to ensure H, regulation performance. It is shown that finding the control gain matrices can be transformed into a standard linear matrix inequality problem and solved via a developed recurrent neural network.