Real-time parallel adaptive neural network control for nonlinear servomechanisms—an approach using direct adaptive techniques

Real-time neural-network-based control of a gyro-mirror line-of-sight stabilization platform

“…In this paper, we further extend the earlier work (Lee et al 1992, Lee and Tan 1993), that we have carried out in neural-net work-based servomechanism control systems and develop a real-time neural-based controller applicable to the gyro-mirror LOS stabilization platform. Note that this is a multivariable servomechanism with nonlinear dynamics and, in the developments presented in this paper, adaptive techniques are utilized and two types of radial basis function (RBF) neural networks are considered: the Gaussian RBF and the Hardy multiquadric RBF.…”

“…Note that these terms arc purposefully combined in such a manner so as to yield Z = Nt -2B as a skew-symmetric matrix; this will subsequently assure that the methods developed previously by Lee and Tan (1993), and can be modified and extended for application to the gyro-mirror platform. With this arrangement, the elements of the matrix Z, in fact, are…”

“…In this section, we shall consider how to develop a neural-network-based controller for the gyro-mirror platform. This will be based on modifications and extensions of the work by Lee and Tan (1993) and . Thus let as.…”

“…Thus, motivated by the neural network-based controllers previously developed by Lee et al (1992), Lee and Tan (1993) and , the controller for the gyro-mirror platform can take the following form: In the design of the neural-network-based controller described above, it is pertinent to point out that the neural networks are only used to compute Mnn(q) and…”

“…However, most of these lose effectiveness when the servomechanisms have fairly severe nonlinear dynamics and, more recently, there has been increasing research and development work to address this problem with the design of control systems, incorporating neural networks. Some examples have been given by Nguyen and Widrow (1990), Ozaki et al (1991), Tzirkel and Fallside (1991), Lee et al (1992), Sanner and Siotine (1992), Lee and Tan (1993) and , but there are undoubtedly other interesting and related developments. In these, neural networks are used in various ways to construct nonlinear control systems to handle a variety of control objectives.…”