• The Transactions of the Korean Institute of Electrical Engineers
• /
• v.39 no.6
• /
• pp.638-648
• /
• 1990

An effective resolved motion control method of redundant manipulators is proposed to minimize the energy consumption and to increase the dexterity while satisfying the physical actuator constraints. The method employs the neural optimization networks, where the computation of Jacobian matrix is not required. Specifically, end effector movement resulting from each joint differential motion is first separated into orthogonal and tangential components with respect to a given desired trajectory. Then the resolved motion is obtained by neural optimization networks in such a way that 1) linear combination of the orthogonal components should be null 2) linear combination of the tangential components should be the differential length of the desired trajectory, 3) differential joint motion limit is not violated, and 4) weighted sum of the square of each differential joint motion is minimized. Here the weighting factors are controlled by a newly defined joint dexterity measure as the ratio of the tangential and orthogonal components.

• Advances in robotics research
• /
• v.1 no.1
• /
• pp.101-126
• /
• 2014

In this article we present modular neural control for a leg-wheel hybrid robot consisting of three legs with omnidirectional wheels. This neural control has four main modules having their functional origin in biological neural systems. A minimal recurrent control (MRC) module is for sensory signal processing and state memorization. Its outputs drive two front wheels while the rear wheel is controlled through a velocity regulating network (VRN) module. In parallel, a neural oscillator network module serves as a central pattern generator (CPG) controls leg movements for sidestepping. Stepping directions are achieved by a phase switching network (PSN) module. The combination of these modules generates various locomotion patterns and a reactive obstacle avoidance behavior. The behavior is driven by sensor inputs, to which additional neural preprocessing networks are applied. The complete neural circuitry is developed and tested using a physics simulation environment. This study verifies that the neural modules can serve a general purpose regardless of the robot's specific embodiment. We also believe that our neural modules can be important components for locomotion generation in other complex robotic systems or they can serve as useful modules for other module-based neural control applications.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.8 no.2
• /
• pp.50-59
• /
• 1998

This paper presents stabilization and adaptive control of flexible single link robot manipulator system by self-recurrent neural networks that is one of the neural networks and is effective in nonlinear control. The architecture of neural networks is a modified model of self-recurrent structure which has a hidden layer. The self-recurrent neural networks can be used to approximate any continuous function to any desired degree of accuracy and the weights are updated by feedback-error learning algorithm. When a flexible manipulator is rotated by a motor through the fixed end, transverse vibration may occur. The motor toroque should be controlled in such a way that the motor rotates by a specified angle, while simultaneously stabilizing vibration of the flexible manipuators so that it is arresed as soon as possible at the end of rotation. Accurate vibration control of lightweight manipulator during the large changes in configuration common to robotic tasks requires dynamic models that describe both the rigid body motions, as well as the flexural vibrations. Therefore, a dynamic models for a flexible single link robot manipulator is derived, and then a comparative analysis was made with linear controller through an simulation and experiment. The results are proesented to illustrate thd advantages and imporved performance of the proposed adaptive control ove the conventional linear controller.

• Journal of Advanced Marine Engineering and Technology
• /
• v.21 no.4
• /
• pp.404-413
• /
• 1997

Direct-drive robots are suitable to position and force control with high accuracy, but it is difficult to design a controller which gives satisfactory perfonnance because of the system's nonlinearity and link-interactions. This paper is concerned with the force control of direct-drive robots. The pro¬posed algorithm consists of feedback controllers and a neural network. Mter the completion of learning, the outputs of feedback controllers are nearly equal to zero, and the neural network con¬troller plays an important role in the control system. Therefore, the optimum adjustment of parameters of feedback controllers is unnecessary. In other words, the proposed algorithm does not need any knowledge of the controlled system in advance. The effectiveness of the proposed algo¬rithm is demonstrated by the experiment on the force control of a parallelogram link-type direct¬drive robot.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.29B no.10
• /
• pp.37-45
• /
• 1992

For a precise manipulator control in the presence of environmental uncertainties, it has long been recognized that the robot should be controlled in a task-referenced space. In this respect, an effective visual servo control system for robot manipulators based on neural networks is proposed. In the proposed control system, a Backpropagation neural network is used first to learn the mapping relationship between the robot's joint space and the video image space. However, in the real control loop, this network is not used in itself, but its first and second derivatives are used to generate servo commands for the robot. Second, and Adaline neural network is used to identify the approximately linear dynamics of the robot and also to generate the proper joint torque commands. Computer simulation has been performed demonstrating the proposed method's superior performance. Futrhermore, the proposed scheme can be effectively utilized in a robot skill acquisition system where the robot can be taught by watching a human behavioral task.

• 제어로봇시스템학회:학술대회논문집
• /
• 1992.10a
• /
• pp.531-536
• /
• 1992

In GMA welding processes, monitoring and control of weld quality are extremely difficult problems. This paper describes a neural network-based method for monitoring and control of weld pool sizes. First, weld pool sizes are estimated via a neural estimator using multi-point surface temperatures, which are strongly related to the formation of weld pool, and then controlled using the estimated pool sizes. Two types of controllers using the pool size estimator are designed and tested. To evaluate the performance of the designed controllers, a series of simulation studies was performed.

• The Transactions of the Korean Institute of Electrical Engineers P
• /
• v.52 no.4
• /
• pp.154-160
• /
• 2003

This paper is proposed a ANN-based rotor position and speed estimation method for IPMSM by measuring the currents. Because the proposed estimator treats the estimated motor speed as the weights, it is possible to estimate motor speed to adapt back propagation algorithm with 2 layered neural network. The proposed control algorithm is applied to IPMSM drive system. The operating characteristics controlled by neural networks are examined in detail.

• International Journal of Internet, Broadcasting and Communication
• /
• v.14 no.4
• /
• pp.173-180
• /
• 2022

In the naked eye observation, the health of livestock can be controlled by the range of activity, temperature, pulse, cough, snot, eye excrement, ears and feces. In order to confirm the health of livestock, this paper uses calf face image data to classify the health status by image shape, color and texture. A series of images that have been processed in advance and can judge the health status of calves were used in the study, including 177 images of normal calves and 130 images of abnormal calves. We used GLCM calculation and Convolutional Neural Networks to extract 6 texture attributes of GLCM from the dataset containing the health status of calves by detecting the image of calves and learning the composite image of Convolutional Neural Networks. In the research, the classification ability of GLCM-CNN shows a classification rate of 91.3%, and the subsequent research will be further applied to the texture attributes of GLCM. It is hoped that this study can help us master the health status of livestock that cannot be observed by the naked eye.

• Transactions on Control, Automation and Systems Engineering
• /
• v.4 no.2
• /
• pp.124-129
• /
• 2002

This paper demonstrates that the largest Lyapunov exponent λ of recurrent neural networks can be controlled efficiently by a stochastic gradient method. An essential core of the proposed method is a novel stochastic approximate formulation of the Lyapunov exponent λ as a function of the network parameters such as connection weights and thresholds of neural activation functions. By a gradient method, a direct calculation to minimize a square error (λ - λ$\^$obj/)$^2$, where λ$\^$obj/ is a desired exponent value, needs gradients collection through time which are given by a recursive calculation from past to present values. The collection is computationally expensive and causes unstable control of the exponent for networks with chaotic dynamics because of chaotic instability. The stochastic formulation derived in this paper gives us an approximation of the gradients collection in a fashion without the recursive calculation. This approximation can realize not only a faster calculation of the gradient, but also stable control for chaotic dynamics. Due to the non-recursive calculation. without respect to the time evolutions, the running times of this approximation grow only about as N$^2$ compared to as N$\^$5/T that is of the direct calculation method. It is also shown by simulation studies that the approximation is a robust formulation for the network size and that proposed method can control the chaos dynamics in recurrent neural networks efficiently.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.15 no.2
• /
• pp.1059-1065
• /
• 2014

Wind turbine system can obtain the maximum wind energy using torque control under the rated wind speed, and wind turbine power is controlled as the rated power using pitch control over the rated wind speed. In this paper, we present a method for wind turbine pitch controller using neural networks. The purpose of the pitch control is to control generator speed and power in the above rated wind speed. To improve the neural network pitch controller, the difference between a rated and current speed of generator has been used for another input of neural networks as well as wind speed. Error back-propagation algorithm is used for training the neural network pitch controller and simulation and Matlab/Simulink is used for verifying that this system is controlled well.