Gradient Descent Control Research Articles

Iterative feedback tuning for internal model controllers

In this paper, the applicability of iterative feedback tuning (IFT) to internal model controllers (IMCs) and smith predictors is examined. IFT is a model-free gradient descent controller tuning tool; refer to Hjalmarsson et al. (IEEE Control Systems Mag. 18 (1998) 26) and Hjalmarsson et al. (Proceedings of the Conference on Decision and Control, Orlando, FL, 1994, pp. 1735) for further details. It is shown that the IFT algorithm can be modified to tune IMCs and, more in particular, Smith predictors. The main difference with the original algorithm is that the cost gradient is obtained by doing four experiments (two recycling experiments) instead of three (one recycling experiment) for conventional controllers.

Microscale adaptive optics: wave-front control with a µ-mirror array and a VLSI stochastic gradient descent controller

The performance of adaptive systems that consist of microscale on-chip elements [microelectromechanical mirror (mu-mirror) arrays and a VLSI stochastic gradient descent microelectronic control system] is analyzed. The mu-mirror arrays with 5 x 5 and 6 x 6 actuators were driven with a control system composed of two mixed-mode VLSI chips implementing model-free beam-quality metric optimization by the stochastic parallel perturbative gradient descent technique. The adaptation rate achieved was near 6000 iterations/s. A secondary (learning) feedback loop was used to control system parameters during the adaptation process, further increasing the adaptation rate.

Chaos Control of Lorenz Model via Direct Gradient Descent Control

This paper is concerned with controlling chaos. Lorenz model, a well known chaotic system, is picked up as a controlled plant in which manipulated variables are additive control inputs. For a controlling technique, we apply a method called the direct gradient descent control(DGDC), which is a dynamic state feedback controller for general nonlinear systems. DGDC manipulates control inputs directly so as to decrease a performance function like the squared error, based on the gradients of the performance function and the steepest descent method. Several simulation results demonstrate the usefulness of the proposed method for the Lorenz chaotic system.

Compact adaptive optical system based on blind optimization and a micromachined membrane deformable mirror

We describe and demonstrate an adaptive optical system based on the combination of a micromachined membrane deformable mirror and the stochastic parallel gradient descent control algorithm. This compact and relatively inexpensive adaptive optical system is used to maximize the coupling of a distorted laser beam into a single-mode optical fiber. The coupling efficiency is improved by 12 dB, and the coupling efficiency after correction is 64% of the diffraction-limited coupling efficiency.

Derivation and Tuning of PID Controller via Output Tracking Direct Gradient Descent Control

Abstract This paper is concerned with derivation and tuning of PID control law. In the first half, output tracking for general nonlinear system is investigated by means of direct gradient descent control (DGDC). In the second half, PID control law for linear systems is derived by using the output tracking DGDC, and at the same time best tuning of PID controller parameters is established. Simulation studies for various plants showed remarkably good performance.

直接勾配降下制御法によるPID制御則の導出と調整

直接勾配降下制御法によるPID制御則の導出と調整

Evaluation of Neural Pattern Classifiers for a Remote Sensing Application

This paper evaluates the classification accuracy of three neural network classifiers on a satellite image-based pattern classification problem. The neural network classifiers used include two types of the Multi-Layer-Perceptron (MLP) and the Radial Basis Function Network. A normal (conventional) classifier is used as a benchmark to evaluate the performance of neural network classifiers. The satellite image consists of 2,460 pixels selected from a section (270 x 360) of a Landsat-5 TM scene from the city of Vienna and its northern surroundings. In addition to evaluation of classification accuracy, the neural classifiers are analysed for generalization capability and stability of results. Best overall results (in terms of accuracy and convergence time) are provided by the MLP-1 classifier with weight elimination. It has a small number of parameters and requires no problem-specific system of initial weight values. Its in-sample classification error is 7.87% and its out-of-sample classification error is 10.24% for the problem at hand. Four classes of simulations serve to illustrate the properties of the classifier in general and the stability of the result with respect to control parameters, and on the training time, the gradient descent control term, initial parameter conditions, and different training and testing sets.