Abstract
The problem of non-affine time-varying nonlinear control systems is addressed in this work through an adaptive state-space neuro-fuzzy control scheme. It combines a eight-layered neuro-fuzzy model to approximate non-affine nonlinear systems' dynamics with a state feedback quadratic stabilising controller. Both the neuro-fuzzy model and controller are updated online within a constrained unscented Kalman filter framework. The proposed generalised state-space neuro-fuzzy model is shown to be an universal approximator, and stability conditions derived for time-varying closed loop systems. Results from a benchmark MIMO system demonstrate the effectiveness of the proposed approach.
Original language | English |
---|---|
Journal | Ieee Transactions On Fuzzy Systems |
DOIs | |
Publication status | Accepted/In press - 1 Jan 2018 |
Keywords
- Kalman filter
- Neuro-Fuzzy Control
- Nonlinear time-varying systems
- recursive learning
- unscented transform