Nonlinear uncertainty observer for AC motor control using the radial basis function networks

Review of recent advancements of direct torque control in induction motor drives – a decade of progress

“…Intelligent controllers were incorporated to improve the performance of sensorless algorithms [91][92][93]97]. Online tuning of controller parameters according to system dynamics was achieved using adaptive supervisory Gaussian-cerebellar model articulation controller (ASGCMAC) [91].…”

“…Several modern control system methods such as SMO, Kalman filter, adaptive systems and intelligent controllers were used to design the flux and speed observers [70, 73–98]. Recently, single sensor was used in three‐phase motor current estimation for cost effective DTC implementation [99–101].…”

“…Steady‐state speed errors observed when PI speed controller is replaced by ANN controllers such as ASGCMAC, supervisory control and cerebellar model articulation controller (CMAC) are recorded in Table 13 [91]. Similarly, fuzzy [49] and ANN [93] were also used to reduce the steady‐state speed error.…”

Neural Network based Robust Adaptive Control for a Variable Stiffness Actuator

“…That could imply high-frequency chattering of controls τ sR (t), due to the behaviour of variables ξ s (t) in (10), that commercial actuators cannot physically support. For that reason, in the following theorem the "σ modification method" [5] will be used to prevent high-frequencies control inputs, while guaranteeing only Uniformly Ultimately Bounded (UUB) stability instead of UAS for the controlled system.…”

“…Recently, NNs are widely used as universal approximators in the area of nonlinear mapping and control problems [5], [6], [8], [9], [13], [14], [10], [11], and among them, Radial-Basis Function Networks (RBFNs) can be noticed due to their increased performance despite of the simple structure, constituted of input, output and hidden layers of normalized Gaussian activation functions [5], [9]. Because RBFNs are universal approximators like fuzzy and neural systems [4], RBFNs have been introduced as a viable solution to the multivariate interpolation problem.…”