基于自适应神经网络的复杂机械臂控制研究

发布时间：2018-01-04 17:41

本文关键词：基于自适应神经网络的复杂机械臂控制研究　出处：《湖南工业大学》2016年硕士论文　论文类型：学位论文

【摘要】：随着科学技术的迅猛发展,机器人技术也得到飞速地发展,并广泛应用于各行各业,如工业、航空航天、军事、医疗等行业。而机械臂是通过模拟人的手臂的一种机械装置,是机器人最主要的执行机构。机械臂系统本身就是一个非线性、强耦合、受干扰的复杂系统。而且,在实践过程中,因工作情况非常复杂,很难建立精确的机械臂系统数学模型,比如负载的不确定,系统参数不确定,甚至完全无模型等情况。另外,从电机输出轴到机械臂的执行轴之间的传动系统不可避免地产生了柔性,所以考虑关节柔性的机械臂控制也成为了当今研究的热点和难点。除此之外,在工业应用中的机械臂出于节约成本或减少测量误差的考虑,一些状态变量是无法测量的,设计观测器也成为了在控制器的设计当中一个重要部分。本论文的研究首先考虑系统参数不确定和部分状态量不可测量的情况下研究出了一种基于一阶滤波观测器的滑模自适应控制器。最后,根据实际过程中的完全无模型的情况下,设计了一种BP神经网络的自适应观测器,并且在此基础上结合神经网络对非线性项的万能逼近原理,采用反演控制的方式实现了机械臂的运动轨迹跟踪。其次,考虑机械臂关节的柔性(特别是关节刚度较小的场合)提出了基于柔性补偿(其本质是增加二次滤波的带宽)的奇异摄动控制方式的机械臂执行端轨迹跟踪控制器。针对完全无模型的情况下,提出了一种RBF神经网络的观测器实现对不可测状态向量的重构,并依此完成了PD控制器的设计。为了实现柔性机械臂的更进一步地智能控制,通过结合神经网络的较强自学习以及联想能力和模糊系统的易于理解推理过程这两者的优点,提出一种模糊神经网络观测器的状态估计和未知非线性项的逼近,进而利用反演控制实现机械臂的轨迹跟踪控制,仿真结果也证明了该控制器的合理性与可行性。最后,通过搭建机械臂控制实验平台,实现DSP目标板与机械臂伺服系统的数据通信,并最终在DSP中执行控制算法并传输控制信号到伺服系统,进而传送动力控制机械臂完成轨迹跟踪的目的。同时,在不中断DSP运行的情况下,采用Matlab下的模块CCSLink实现DSP、CCS和matlab实时交互数据。最终,通过实验结果更进一步验证了控制策略的可行性与有效性。
[Abstract]:With the rapid development of science and technology, robot technology has also been rapidly developed, and widely used in various industries, such as industry, aerospace, military. The mechanical arm is a kind of mechanical device which simulates the human arm and is the most important actuator of the robot. The robot arm system itself is a nonlinear and strong coupling. In practice, it is very difficult to establish accurate mathematical model of manipulator system, such as uncertain load and uncertain system parameters. In addition, the transmission system from the motor output shaft to the actuator shaft of the manipulator inevitably produces flexibility. Therefore, the control of manipulator with flexible joints has become a hot and difficult point. In addition, in industrial applications, the robot arm is considered to save cost or reduce the measurement error. Some state variables are unmeasurable. The design of observer has become an important part of controller design. In this paper, first of all, considering the uncertainty of system parameters and the unmeasurable state of part of the system, a new method based on first-order filter observation is proposed. The sliding mode adaptive controller. Finally. In this paper, an adaptive observer of BP neural network is designed, and the universal approximation principle of neural network to nonlinear term is combined. The motion trajectory tracking of the manipulator is realized by inverse control. Secondly. Considering the flexibility of the manipulator joints (especially when the stiffness of the joints is small), a new method based on flexibility compensation is proposed (the essence of which is to increase the bandwidth of the secondary filtering). A singularly perturbed control method for the manipulator actuator trajectory tracking controller. For the case of no model. An observer based on RBF neural network is proposed to reconstruct the unmeasurable state vector, and the PD controller is designed in order to realize further intelligent control of the flexible manipulator. By combining the strong self-learning of neural networks and the advantages of associative ability and easy to understand the reasoning process of fuzzy systems, a fuzzy neural network observer state estimation and approximation of unknown nonlinear terms are proposed. Then the trajectory tracking control of the manipulator is realized by inverse control. The simulation results also prove the rationality and feasibility of the controller. Finally, the manipulator control experimental platform is built. The data communication between the DSP target board and the manipulator servo system is realized. Finally, the control algorithm is implemented in DSP and the control signal is transmitted to the servo system. At the same time, under the condition of not interrupting the DSP running, the module CCSLink under Matlab is used to realize the DSP. CCS and matlab interact with each other in real time. Finally, the feasibility and effectiveness of the control strategy are further verified by the experimental results.
【学位授予单位】：湖南工业大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP241;TP183

【参考文献】