A PID-like neural network control method for a 5PUS-RPUR parallel robot considering force coupling errors

In this paper, a neural network control based on optimal quadratic regulators is developed for the stabilization of constrained nonlinear robotic systems. This method of robotic control is performed by adding an optimal control, generated from the dynamics of the position error, to a neural control, estimated through a three layers neural network. Solving an algebraic Riccati equation, solutions of the Hamilton Jacobi Bellman (HJB) equation are found for the dynamic error optimal control. The adaptation algorithm of the neural control is derived from the Lyapunov analysis for the robotic system overall stability. The simulations results of a six degrees of freedom arm manipulator, Puma 560, with this neural optimal control, are presented to validate the proposed approach. Global stability, of the constrained robot, is assured through the developed controller, in the presence of uncertain gravity vector, viscous and Coulomb's friction and external disturbances.

Single-phase grid-connected inverters are widely used to connect small-scale distributed renewable resources to the grid. However, unlike a three-phase system, control for a single-phase inverter is more challenging, especially when the inverter is used with an LCL filter. This paper proposes a novel recurrent neural network-based vector control method for a single-phase inverter with an LCL filter. The neural network is trained based on adaptive dynamic programming principle, and the objective of the training is to approximate optimal control. The Levenberg–Marquardt plus forward accumulation through time algorithm is developed for training the proposed recurrent neural network controller. The neural network vector control approach is compared with the conventional control methods, including the conventional PI-based vector control method and the PR-based control technique for single-phase inverters. Both the simulations and hardware experiments demonstrate the great advantages of the proposed neural network vector control over the conventional control methods. Compared with conventional control methods, the neural network control allows for low sampling rate and low switching frequency, while maintaining high performance in controlling a single-phase inverter. In addition, no specific damping policy is required to implement the proposed neural network vector control for an LCL -filter based single-phase inverter. The study shows that the neural network vector control is a robust control method, and can provide better control performance even when facing system parameter changes, while under this case, both the conventional PI-based vector control and the PR-based control failed to yield the acceptable results.

Sliding mode adaptive neural network control for hybrid visual servoing of underwater vehicles

Vision-admittance-based adaptive RBFNN control with a SMC robust compensator for collaborative parallel robots

This paper considers the trouble of the usage of approximate strategies for realizing the neural controllers for nonlinear SISO systems. In this paper, we introduce the nonlinear autoregressive moving average (NARMA-L2) model which might be approximations to the NARMA model. The nonlinear autoregressive moving average (NARMA-L2) model is an precise illustration of the input–output behavior of finite-dimensional nonlinear discrete time dynamical systems in a neighborhood of the equilibrium state. However, it isn't always handy for purposes of neural networks due to its nonlinear dependence on the manipulate input. In this paper, nerves system based arm position sensor device is used to degree the precise arm function for nerve patients the use of the proposed systems. In this paper, neural network controller is designed with NARMA-L2 model, neural network controller is designed with NARMA-L2 model system identification based predictive controller and neural network controller is designed with NARMA-L2 model based model reference adaptive control system. Hence, quite regularly, approximate techniques are used for figuring out the neural controllers to conquer computational complexity. Comparison were made among the neural network controller with NARMA-L2 model, neural network controller with NARMA-L2 model system identification based predictive controller and neural network controller with NARMA-L2 model reference based adaptive control for the preferred input arm function (step, sine wave and random signals). The comparative simulation result shows the effectiveness of the system with a neural network controller with NARMA-L2 model based model reference adaptive control system. Index Terms--- Nonlinear autoregressive moving average, neural network, Model reference adaptive control, Predictive controller DOI: 10.7176/JIEA/10-3-03 Publication date: April 30 th 2020

Research on Structure Design and Control Method of Underwater Exploration Robot

Home robots have the potential to assist older adults in maintaining their independence. However, robots deployed in older adults' homes will be required to interact with untrained, novice users. The way untrained users, such as older adults, provide commands or control the robot (i.e., "method of robot control") will likely impact the ease of use and adoption of the robot. The current study explored older adults' preferences for controlling robots. Twelve independently-living older adults (ages 68-79) observed a functioning personal robot in a home setting, and were interviewed about their opinions regarding specific methods of robot control (i.e., laser pointer, physical manipulation, and devices). The older adults perceived advantages and disadvantages of these specific methods, including 'specificity in command', 'accurate robot performance', 'limitations in their own physical capability', and 'challenges in using control device.' The older adults also completed a questionnaire measuring their willingness to use 10 different types of methods of robot control. These data revealed that older adults were willing to use a variety of methods. Although older adults were limited in their spontaneous ideas about robot control (i.e., limited to voice command), once exposed to other options they were willing and open to a variety of control methods.

In this paper, a direct adaptive neural network control (DANNC) method is developed to deal with the multi-variable (dissolved oxygen concentration and nitrate concentration) tracking control problem in wastewater treatment processes (WWTPs), which avoids the perplex issue of establishing the plant model of WWTP and has the excellent adaptive ability. The DANNC system is composed of neural controller and compensation controller. The neural controller is employed to approximate an ideal control law, and the compensation controller is designed to offset the network approximation error. The controller parameters' adaptive laws are deduced by the Lyapunov theorem. Simulation results, based on the international benchmark simulation model No.1 (BSM1), show that the control accuracy and dynamic performance of the DANNC method are improved nicely.

Advanced flexible endoscopes and instruments with multiple degrees of freedom enable physicians to perform challenging procedures such as the removal of large sections of mucosal tissue. However, these advanced endoscopes are difficult to control and require several physicians to cooperate. In this article, we present a robotic system that allows the physician to control an instrument in an intuitive way, using a haptic device. Performance with the robotic and conventional control methods were compared in a human subjects experiment. Subjects used both methods to tap a series of targets. They performed four trials while looking at the endoscopic monitor, and two trials while looking at the instrument directly. Subjects were significantly faster using the robotic method, 54 s vs 164 s. Their performance in the second trial was significantly improved with respect to the first trial. This study provides evidence that the robotic control method can be implemented to improve the performance of physicians using advanced flexible endoscopes.

The realization and capability of magnetic bearing principally depends on the design of controller. It is difficult to induce its precise mathematic model because the magnetic bearing has complex non-linearity. The classical PID control method focus on systems having precise mathematic models. The neural network control method does not need the precise mathematic model, and has entirely different information processing approach compared to the classical PID control. The neural network, based on the principles of self-adaptive and being-trained, has self-study capability, so it adapts to controlling a magnetic bearing system. In this paper, we simulate both the neural network PID control algorithm and the classical PID control algorithm with the disturbances of output force exist, and conclude that the neural network PID control is superior to the classical PID control in respect of adjusting time and overshooting values.

Taking into account the existing neural network control method of large amount of calculation and application is not strong, in this paper, a complex neural network control structure is designed for the nonlinear and strong coupling system, which is coupled by the N sub neural network, at the same time, the paper proposes a neural network control method with end point bias. The method assigns different weights to each sub item of the mean variance objective function, so that the determination of neural network parameters is effectively biased towards the end point error value, so as to ensure the terminal stability of the system. Finally, the applicability of the proposed method is verified by a nonlinear strong coupling system with multiple inputs and multiple outputs, and the expected results are obtained.

Unmanned vehicle intelligent control methods are advantaged in this paper. For path control of an unmanned vehicle, tracking method is proposed based on neural network. A neural network is made through experiments. Neural network's input are velocity, friction coefficient, hope radius, output is velocity difference. Then prevision control method is used to steering control. This neural network control method can adapt different velocity, ground surface and turning radius. Control method is simple and reliable. For steering control of wheeled mobile robots with complex mathematical models, a multistep neural network is proposed. Neural networks learn speed, maximum overshoot, overshoot time and steady steering angle in different cases in a reduced learning capacity. As for turning control of wheeled robots, fuzzy neural network model and GA (genetic algorithm) PID control method can be used. Fuzzy GA PID control algorithm is simple, and efficiency of PID parameters can be judged directly. A GA fuzzy neural network is used for steering control of wheeled mobile robots. At first, a neural network model of mobile robot is established. Then, a fuzzy neural network controller is constructed, and GA method is used to find best control parameters. Combining direction and speed control of wheeled mobile robots, GA fuzzy neural networks are used. At first, a fuzzy neural network controller is built, then, a GA optimum algorithm is used to find best parameters for controller. All methods and algorithms proposed in this paper are simulated and tested. Simulation and experiment results show that it is efficient and reliable.

In this paper, an adaptive neural network (NN) control method is proposed for the problem of nonlinear course control of ships with input constraints and unknown direction control gains. Specifically, dynamic surface control is used to overcome the problem of explosion of complexity inherent in the backstepping technique, and the Nussbaum function is employed to deal with the unknown signs of control gains. It is proved that the proposed adaptive NN control method, which is composed of dynamic surface control and a backstepping technique with the Nussbaum gain function, is able to guarantee uniform ultimate boundedness of all the signals in the controlled system. In addition, the tracking error between the output of the controlled system and a desired trajectory is shown to converge to a small neighbourhood of the origin. Finally, one example is introduced to illustrate the proposed theoretical results.

A single-phase Cascaded H-Bridge (CHB) grid-tied multilevel inverter is introduced with a detailed discussion of the proposed novel neural controller for better efficiency and power quality in the integration of renewable sources. An LCL (inductor-capacitor-inductor) filter is used in the multilevel inverter system to achieve better harmonic attenuation. The proposed Neural Network (NN) controller performs the inner current control and tracks the references generated from the outer loop to satisfy the requirements of voltage or power control. Two multicarrier-based Pulse Width Modulation (PWM) techniques (phase-shifted modulation and level-shifted modulation) are adopted in the development of the simulation model to drive the multilevel inverter system for the evaluation of the neural control technique. Simulations are carried out to demonstrate the effectiveness and efficient outcomes of the proposed neural network controller for grid-tied multilevel inverters. The advantages of the proposed neural control include a faster response speed and fewer oscillations compared with the conventional Proportional Integral (PI) controller based vector control strategy. In particular, the neural network control technique provides better harmonics reduction ability.

Since a robotic manipulator has a complicated mathematical model, it is difficult to design a control system based on the complicated multi-variable nonlinear coupling dynamic model. Intelligent controllers using fuzzy and neural network approaches do not need a real mathematical model to design the control structure and have attracted the attention of robotic control researchers recently. A traditional fuzzy logic controller does not have learning capability and it needs a lot of effort to search for the optimal control rules and the shapes of membership functions. Owing to the time-varying behaviour of the system, the required fine tracking accuracy is difficult to achieve by adjusting the fuzzy rules only. The implementation problems of neural network control are the initial training and initial transient stability. In order to improve the position control accuracy and system robustness for industrial applications, a neural controller is first trained off-line by using the input and output (I/O) data of a traditional fuzzy controller. Then the neural controller is implemented on a five-degrees-of-freedom robot with a back propagation algorithm for online adjustment. The experimental results show that this neural network controller achieved the required trajectory tracking accuracy after 15 on-line operations.