A neural approach for control of nonlinear systems with feedback linearization

In this paper several schemes for feedback linearization using neural networks have been investigated and compared, Then an approach to design a neurocontroller in the sense of feedback linearization is introduced. The contents include: 1) full input-output linearization when a system has relative degree n; 2) partial input-output linearization when a system has relative degree r (r < n); and 3) approximate linearization when the involutivity condition does not hold. Corresponding programs and examples are given to illustrate the proposed methodology.