Control with Dual Mode Adaptation Using a RBF Network
Adaptive Control, Dual Mode Control, Neural Network, Radial Basis Functions
In this work, a RBF network with gaussians is adapted through a combination of switching and integral laws. The idea is taking the best features of each method in the control of unknown nonlinear system. A hyperbolic tangent is used instead of signal function what reduces the chattering phenomenon. A hyperbolic secant is used to regulate the integral law, increasing its effects on steady state and reducing on transient time. This method produces a smooth control signal with a significant chattering reduction or even its elimination and fast and less oscillatory system output. The process is in real time, and the error between the reference model and system output is the controlled variable. Only measurements of input and output of the plant is needed. Knowledge about parameters and disturbances is unnecessary too. It is presented a Lyapunov proof for dual mode method and comparisons through simulations.