气动肌肉碰撞感知及其关节转角控制

发布时间：2018-04-17 01:19

本文选题：机器人 + 机械手臂　；参考：《中国计量学院》2015年硕士论文

【摘要】：机器人不仅要有较高的控制精度,还需要具备碰撞感知能力和柔顺性。用气动肌肉驱动的机器人具有质量轻、柔顺性和仿生性好等优点,应用前景广泛。但是气动肌肉具有很多复杂特性,对气动肌肉建模、控制以及在其表面安装传感器困难。本文主要研究了两个问题：一是根据气动肌肉两端的差压信号感知气动肌肉径向碰撞；二是研究气动肌肉关节的控制算法。建立了气动肌肉特性测试平台,设计单自由度气动肌肉关节,测试了气动肌肉和关节特性。采用二阶高斯函数和三阶傅立叶函数对原始曲线和迟滞性曲线拟合,得到气动肌肉长度-气压模型；测试结果表明气动肌肉及其驱动的关节具有很强的非线性和迟滞性,而且会随着气压、负载变化而变化。提出根据气动肌肉两端差压信号感知气动肌肉径向碰撞(冲击)的方法,并与轴向冲击区分；基于流体阻抗法,建立了轴、径向冲击差压信号模型。搭建实验平台和数据采集系统,通过实验研究了负载、冲击作用、内部气压、碰撞位置对差压信号的影响；实验结果表明相同条件下径向冲击作用产生的差压信号幅值大于轴向,轴向相频曲线变化杂乱,径向相频曲线有周期性变化规律。四种因素会改变差压信号的幅值,但不会改变相频曲线的变化规律。采用自相关函数法提取相频曲线的周期性特征,实现了轴、径向冲击区分。设计了改进的神经元PID控制算法,用Sigmoid函数定义神经元PID控制算法的增益系数,提高了控制算法的自适应性,并在参数学习过程中添加衰减因子来提高参数学习的收敛速度。针对气动肌肉充、放气过程中表现出来的迟滞性,设计了双层结构的MFA-DSCMAC (Model Free Adaptive-Double Structure Cerebellar Model Articulation Controller)控制算法,对充、放气过程分别进行学习,从而有效地补偿了迟滞性；设计误差可信度评估函数来调整学习率,抑制突发干扰对神经网络学习的破坏。搭建了一个五自由度机械手臂,基于工控机和C++软件建立控制系统,采用串联的关节控制算法控制机械手臂完成倒水作业。
[Abstract]:The robot should not only have high control precision, but also have collision sensing ability and flexibility.The robot driven by pneumatic muscle has many advantages, such as light weight, flexibility and good bionics.But pneumatic muscle has many complex characteristics, it is difficult to model, control and install sensors on the surface of pneumatic muscle.This paper mainly studies two problems: one is to perceive radial collision of pneumatic muscle according to the differential pressure signal at both ends of pneumatic muscle; the other is to study the control algorithm of pneumatic muscle joint.The pneumatic muscle characteristic test platform is established, and the single degree of freedom pneumatic muscle joint is designed, and the pneumatic muscle and joint characteristics are tested.By using the second-order Gao Si function and the third-order Fourier function to fit the original curve and the hysteresis curve, the length pressure model of the pneumatic muscle is obtained, and the test results show that the pneumatic muscle and its actuated joints have strong nonlinearity and hysteresis.And it changes with the pressure and the load.A method for sensing radial impact (shock) of pneumatic muscle based on differential pressure signals at both ends of pneumatic muscle is proposed, which is distinguished from axial impact, and a model of differential pressure signal for axial and radial impact is established based on fluid impedance method.The experiment platform and data acquisition system are built, and the effects of load, impact, internal pressure and collision position on differential pressure signal are studied experimentally.The experimental results show that the amplitude of differential pressure signal produced by radial shock is larger than that of axial direction, the axial phase frequency curve is chaotic, and the radial phase frequency curve has periodic variation rule.Four factors will change the amplitude of differential pressure signal, but will not change the law of phase frequency curve.The periodic characteristic of phase frequency curve is extracted by autocorrelation function method.The improved neural PID control algorithm is designed, and the gain coefficient of the neural PID control algorithm is defined by Sigmoid function, which improves the self-adaptability of the control algorithm, and the attenuation factor is added in the process of parameter learning to improve the convergence rate of the parameter learning.Aiming at the hysteresis in the process of pneumatic muscle filling and exhalation, a double-layer MFA-DSCMAC model Free Adaptive-Double Structure Cerebellar Model Articulation controller is designed, which can be used to study the process of filling and discharging respectively, thus compensating the hysteresis effectively.The error reliability evaluation function is designed to adjust the learning rate and to suppress the damage to the learning of neural network caused by burst interference.A control system based on industrial computer and C software is set up for a five-degree-of-freedom manipulator, and a series joint control algorithm is used to control the robot arm to complete the operation of pouring water.
【学位授予单位】：中国计量学院
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TH138;TP242

【相似文献】