Robot learning control based on recurrent neural network inverse model

A non-linear model-based feedforward, feedback, and learning controller is presented. This controller can control a non-linear plant such as a robot whose dynamics are initially unknown. In the feedforward part, a recurrent neural network (RNN) is used to model the inverse dynamics of the plant. In the feedback part, a PD controller is added to handle unmodeled dynamics and disturbances. Furthermore, an add-on learning controller is established to reduce tracking errors for repetitive tasks. The controller is validated with the control of a simulated two-joint manipulator. Simulation results show that the controller can successfully learn the inverse dynamics of a robot, perform accurate tracking for a general trajectory, and improve its own performance over the repetitions of a trajectory, with and without a payload change. © 1997 John Wiley & Sons, Inc.     

基于神经网络的机器人轨迹鲁棒跟踪控制

在神经网络辨识的基础上 ,提出一种新的鲁棒迭代学习控制方法。该方法利用神经网络对非线性系统进行在线辨识 ,产生迭代学习控制算法的前馈作用 ,并与实时反馈控制相结合 ,实现连续轨迹跟踪控制。仿真结果表明 ,该方法能克服机器人系统动力学模型的不确定性和外部干扰 ,且以极少的学习次数和网络训练次数达到满意的跟踪控制要求 ,具有良好的鲁棒性和控制性能     

基于模糊神经网络的滑模控制

研究了一类不确定性非线性系统的滑模变结构控制,提出了一种基于模糊神经网络（Ｆｕｚｚｙ Ｎｅｕｒａｌ Ｎｅｔｗｏｒｋｓ）的滑模变结构设计方法,设计了控制器的结构,利用动态反向传播算法实现滑模控制,这种方法与一般变结构控制相比不但具有强的鲁棒性而且还能有效地消除抖动现象,同时在设计中不需要知识系统中不确定性和扰动的上界,另外还运用Ｌｙａｐｕｎｏｖ函数从理论上分析上了系统的稳定性。仿真结果说明了本文所提     

人工神经网络在机械手动力学辨识和位置控制中的应用

本文提出了用人工神经网络近似机械手的逆动力学模型,实现基于模型的非线性控制方案,并以实际的两臂液压机械手为对象给出了仿真结果,整个方案的实现不需要任何关于系统模型的知识.仿真结果表明所研究的位置控制系统具有良好的跟踪性能,并且展示了人工神经网络解决非线性系统辨识和控制问题的潜力.     

基于神经网络的机器人轨迹跟踪控制

针对机器人模型未知情况，讨论了用神经网络和反馈控制实现机械手的跟踪控制。提出一种基于参考误差的投影算法来训练网络权值，训练后网络输出能逼近期望的前馈力矩，并从理论上证明跟踪误差的收敛性。仿真结果表明方案具有较好的跟踪性能和较强的抗干扰能力。     

移动焊接机器人焊缝跟踪控制研究及仿真

为了解决轮式移动焊接机器人的焊缝跟踪问题,文中从移动焊接机器人的运动学和动力学模型出发,采用分段运动学到动力学的方法设计焊缝跟踪控制器。控制算法结合积分Backstepping算法和单层神经元网络控制算法,利用单层神经元网络的自学习和自适应能力克服机器人模型参数部分未知和扰动的影响,使跟踪更加快速、平滑。经MATLAB仿真验证了该控制算法的有效性。     

Adaptive output feedback tracking control of robot manipulators using position measurements only

In this paper, a new adaptive neuro controller for trajectory tracking is developed for robot manipulators without velocity measurements, taking into account the actuator constraints. The controller is based on structural knowledge of the dynamics of the robot and measurements of joint positions only. The system uncertainty, which may include payload variation, unknown nonlinearities and torque disturbances is estimated by a Chebyshev neural network (CNN). The adaptive controller represents an amalgamation of a filtering technique to generate pseudo filtered tracking error signals (for the elimination of velocity measurements) and the theory of function approximation using CNN. The proposed controller ensures the local asymptotic stability and the convergence of the position error to zero. The proposed controller is robust not only to structured uncertainty such as payload variation but also to unstructured one such as disturbances. Moreover the computational complexity of the proposed controller is reduced as compared to the multilayered neural network controller. The validity of the control scheme is shown by simulation results of a two-link robot manipulator. Simulation results are also provided to compare the proposed controller with a controller where velocity is estimated by finite difference methods using position measurements only.     

Fault detection in catalytic cracking converter by means of probability density approximation

The paper deals with a model-based fault diagnosis for a catalytic cracking converter process realized using artificial neural networks. Modelling of the considered process is carried out by using a locally recurrent neural network. Decision making about possible faults is performed using statistical analysis of a residual. A neural network is applied to density shaping of a residual. After that, assuming a significance level, a threshold is calculated. The proposed approach is tested on the example of a catalytic cracking converter at the nominal operating conditions as well as in the case of faults.     

Creating efficient nonlinear neural network process models that allow model interpretation

The KBANN (knowledge-based artificial neural networks) approach uses neural networks to refine knowledge that can be written in the form of simple propositional rules. This idea is extended by presenting the MANNIDENT (multivariable artificial neural network identification) algorithm by which the mathematical equations of linear process models determine the topology and initial weights of a network, which is further trained using backpropagation. This method is applied to the task of modelling a non-isothermal CSTR in which a first-order exothermic reaction is occurring. This method produces statistically significant gains in accuracy over both a standard neural network approach and a linear model. Furthermore, using the approximate linear model to initialize the weights of the network produces statistically less variation in accuracy. By structuring the neural network according to the approximate linear model, the model can be readily interpreted.     

Tracking control of a nonholonomic mobile robot using compound cosine function neural networks

The purpose of this paper is to propose a compound cosine function neural network with continuous learning algorithm for the velocity and orientation angle tracking control of a nonholonomic mobile robot with nonlinear disturbances. Herein, two neural network (NN) controllers embedded in the closed-loop control system have the simple continuous learning and rapid convergence capability without the dynamics information of the mobile robot to realize the adaptive control of the mobile robot. The neuron function of the hidden layer in the three-layer feed-forward network structure is on the basis of combining a cosine function with a unipolar sigmoid function. The developed neural network controllers have simple algorithm and fast learning convergence because the weight values are only adjusted between the nodes in hidden layer and the output nodes, while the weight values between the input layer and the hidden layer are one, i.e. constant, without the weight adjustment. Therefore, the main advantages of this control system are the real-time control capability and the robustness by use of the proposed neural network controllers for a nonholonomic mobile robot with nonlinear disturbances. Through simulation experiments applied to the nonholonomic mobile robot with the nonlinear disturbances which are considered as dynamics uncertainty and external disturbances, the simulation results show that the proposed NN control system of nonholonomic mobile robots has real-time control capability, better robustness and higher control precision. The compound cosine function neural network provides us with a new way to solve tracking control problems for mobile robots.     

一个基于神经网络模型的鲁棒自适应控制算法

对于具有不确定因素的离散非线性动态系统,通过校正神经网络预报器的输出,运用加权 预报控制性能指标和网络辨识器模型局部线性化的思想,提出了一个间接鲁棒自适应神经网 络控制算法,仿真研究证实了该控制策略的鲁棒性和有效性.     

Neural Approximation-based Model Predictive Tracking Control of Non-holonomic Wheel-legged Robots

This paper proposes a neural approximation based model predictive control approach for tracking control of a nonholonomic wheel-legged robot in complex environments, which features mechanical model uncertainty and unknown disturbances. In order to guarantee the tracking performance of wheel-legged robots in an uncertain environment, effective approaches for reliable tracking control should be investigated with the consideration of the disturbances, including internal-robot friction and external physical interactions in the robot’s dynamical system. In this paper, a radial basis function neural network (RBFNN) approximation based model predictive controller (NMPC) is designed and employed to improve the tracking performance for nonholonomic wheel-legged robots. Some demonstrations using a BIT-NAZA robot are performed to illustrate the performance of the proposed hybrid control strategy. The results indicate that the proposed methodology can achieve promising tracking performance in terms of accuracy and stability.

     

A neural network model-based observer for idle speed control of ignition in SI engine

The paper presents an algorithm of idle speed stabilization in the spark ignition automotive engine by means of spark advance control. The algorithm is based on a well-known approach of a model-based adaptive control and uses artificial neural networks. The control algorithm is based on a neural network model observer of the additional effective torque. The additional load is estimated as difference between effective torque, estimated by the neural network observer, and brake torque, estimated on the basis of a linear quadratic model. In that case the additional load is understood as the sum of the alternator brake torque (additional load form electric car equipments) and the momentary and/or permanent changes of the engine’s characteristics.On the basis of estimated values of the additional load, the required value of angular acceleration is determined to make the engine return to the specified speed. This acceleration is achieved by adjusting the spark advance. The required value of spark advance is estimated by means of a neural network model converse to that of the effective torque.The algorithm was experimentally compared with PID and adaptive algorithms in the same test bed. The tests were conducted under sudden change of external load. The proposed algorithm proved to be more effective in terms of control error.     

Adaptive actor-critic learning for the control of mobile robots by applying predictive models

In this paper, we propose two methods of adaptive actor-critic architectures to solve control problems of nonlinear systems. One method uses two actual states at time k and time k+1 to update the learning algorithm. The basic idea of this method is that the agent can directly take some knowledge from the environment to improve its knowledge. The other method only uses the state at time k to update the algorithm. This method is called, learning from prediction (or simulated experience). Both methods include one or two predictive models, which are assumed to be applied to construct predictive states and a model-based actor (MBA). Here, the MBA as an actor can be viewed as a network where the connection weights are the elements of the feedback gain matrix. In the critic part, two value-functions are realized as a pure static mapping, which can be reduced to a nonlinear current estimator by using the radial basis function neural networks (RBFNNs). Simulation results obtained for a dynamical model of nonholonomic mobile robots with two independent driving wheels are presented. They show the effectiveness of the proposed approaches for the trajectory tracking control problem.     

基于神经网络与粒子滤波的柔性臂控制方法研究

基于奇异摄动法将单连杆柔性臂系统分解为慢变、快变子系统,采用混合控制方法;设计了基于粒子滤波的神经网络控制器来线性化慢子系统,使其跟踪期望轨迹;采用粒子滤波训练神经网络克服了BP算法收敛速度慢、易陷入局部极小值的缺陷,及扩展卡尔曼滤波方法带来的模型线性化损失;对于快变系统采用最优控制方法;仿真结果表明:在神经网络训练误差收敛速度及精度方面,粒子滤波要比BP及卡尔曼滤波要好;组合控制方法能有效地抑制柔性臂弹性振动,轨迹跟踪迅速准确,精度方面也是前者最优。     

Applications of adaptive neural network control to an unmanned airship

This paper represents an application of a neural network-based adaptive control to the Stability and Control Augmentation System(SCAS) of an unmanned airship whose maneuvers consist of diverse flight phases at low speeds. The neural network (NN) based adaptive SCAS is based on the inversion of a linear model of the airship at a nominal operating point and the adaptation of neural networks to unmodeled dynamics, parameter variations, and uncertain environments. This paper also presents an evaluation of the adaptive SCAS with flight test results and simulation results. In this evaluation, an outer-loop control is used. The autopilot is designed using a classical PID control algorithm for trajectory line tracking and altitude hold modes. Moreover, the adaptive SCAS approach showed superiority over the classical PID design approach in terms of the gain tuning process during a flight test.     

Neural network MBPC for mobile robot path tracking

This paper presents a way of implementing a model-based predictive controller (MBPC) for mobile robot path tracking. The method uses a non-linear model of mobile robot dynamics and thus allows an accurate prediction of the future trajectories. Constraints on the maximum attainable speeds are also considered by the algorithm. A multilayer perceptron is used to implement the MBPC. The perceptron has been trained to reproduce the MBPC bahaviour in a supervised way. Experimental results obtained when applying the neural network controller to a TRC labmate mobile platform are given in the paper.     

Control of a nonholomic mobile robot: Backstepping kinematics into dynamics

A dynamical extension that makes possible the integration of a kinematic controller and a torque controller for nonholonomic mobile robots is presented. A combined kinematic/torque control law is developed using backstepping, and asymptotic stability is guaranteed by Lyapunov theory. Moreover, this control algorithm can be applied to the three basic nonholonomic navigation problems: tracking a reference trajectory, path following, and stabilization about a desired posture. The result is a general structure for controlling a mobile robot that can accommodate different control techniques, ranging from a conventional computed-torque controller, when all dynamics are known, to robust-adaptive controllers if this is not the case. A robust-adaptive controller based on neural networks (NNs) is proposed in this work. The NN controller can deal with unmodeled bounded disturbances and/or unstructured unmodeled dynamics in the vehicle. On-line NN weight tuning algorithms that do not require off-line learning yet guarantee small tracking errors and bounded control signals are utilized. © 1997 John Wiley & Sons, Inc.     

Performance of a neural adaptive tracking controller formulti-input nonlinear dynamical systems in the presence of additive andmultiplicative external disturbances

We discuss the tracking problem in the presence of additive and multiplicative external disturbances, for affine in the control nonlinear dynamical systems, whose nonlinearities are assumed unknown. Based on a recurrent high order neural network (RHONN) model of the unknown plant, a smooth control law is designed to guarantee the uniform ultimate boundedness of all signals in the closed loop. Certain measures are utilized to test its performance. The controller, which can be viewed as a nonlinear combination of three high order neural networks, does not require knowledge regarding upper bounds on the optimal weights, modeling error and external disturbances. Simulations performed on a simple example illustrate the approach