具有执行器非线性和状态约束的机器人自适应控制

发布时间：2018-08-20 09:12

【摘要】：随着信息、机械、材料等多学科的交叉发展,机器人在辅助甚至替代人类进行协调工作方面崭露出良好的应用潜质和强劲的市场需求。因此,机器人技术的研究不仅具有明显的应用前景,而且具有重要的理论价值。本文将机器人操作过程描述为多自由度运动/力混合非线性系统,重点围绕如何处理执行器非线性和状态约束等不确定性因素展开研究,旨在提高机器人控制性能。本文结构安排如下。第一章阐述相关研究背景及研究意义。第二章概述机器人系统建模和控制在内的背景知识。第三章至第七章共计五个章节,是本文的主要内容,包括三方面：1)、第三章至第四章主要研究具有执行器非线性的机器人自适应模糊协调控制；2)、第五章研究基于Nussbaum函数方法的广义执行器非线性机器人自适应控制；3)、第六章至第七章主要研究具有状态约束的机器人自适应神经网络协调控制。具体来说,上述五章依次对应如下内容：一、研究了在执行器间隙非线性情况下机器人抓取物体的协调控制问题,提出了一种补偿执行器间隙的运动/力分散模糊控制方案。首先,基于间隙非线性逆补偿思想,构造了一种执行器间隙逆模型自适应控制方法。然后,通过建立分散鲁棒自适应模糊协调控制方法,保证了物体运动和内力分别收敛于期望值。最后,将所提方法在双臂机器人系统上进行仿真计算并与已有方法进行比较,结果证实了所提方法的有效性。二、研究了多机械臂在执行器磁滞非线性和运动受限情况下的协调控制问题,提出了一种基于Barrier李雅普诺夫函数方法的机器人自适应模糊控制方案。首先,在多机械臂的执行器动力学方程中建立了磁滞模型,进而引入了自适应控制技术补偿并减少来自未知磁滞非线性的影响。紧接着,将不同工作环境下运动要求考虑进控制器设计,有效地限制了被操作物体的运动范围。并且,基于李雅普诺夫稳定性定理,保证了所提方法在多机械臂协调过程中运动和力控制性能。最后,多组比较结果说明了所提方法的有效性。三、研究了具有广义未知执行器非线性的机器人系统运动控制方法,建立了新型的Nussbaum函数分析工具,解决了未知时变非线性控制系数的稳定性分析问题。相比较于已有的Nussbaum函数方法,所提的方法不仅将多个未知控制系数从常数扩展为时变量,而且去除了多个控制系数上下界已知的假设条件。将自适应鲁棒控制与Nussbaum函数方法相结合,保证了机器人系统状态在广义执行器非线性情况下渐近收敛到期望轨迹。进一步地,为减少因使用传统Nussbaum函数引起的控制抖动,提出了一种基于饱和Nussbaum函数的控制方法。所提Nussbaum函数是基于时间扩展思想构造,减少了传统幅值扩展Nussbaum函数带来的值域不确定问题。并且,通过与自适应控制方法相结合,建立了一种处理多个未知时变控制系数的控制方法,促成了针对多输入多输出系统的稳定性分析,保证了机器人系统在未知执行器动态情况下实现运动状态渐近跟踪期望轨迹的控制目标。四、研究了双机械臂协调控制过程中由输出机构导致的状态磁滞约束问题,提出了一种自适应神经网络控制器,实现了机器人运动/力协调控制。所提方法完成了输出磁滞约束特性分析,并应用Nussbaum函数方法解决了由反推控制设计方法带来的时变系数分析困难。同时,结合自适应神经网络控制方法,实现了神经网络权重矩阵的上界估计,减少了待更新自适应律数量,降低了完成实时控制所需的计算负荷。最后,将所提控制方法与传统方法进行性能比较与评估,其结果进一步验证了所提方法的有效性、优越性和鲁棒性。五、研究了具有未知输出死区约束和物体不确定性的多机械臂协调控制问题,提出了一种多机械臂运动/力自适应神经网络协调控制方法。给出了一种死区建模方法,解决了反推控制设计过程中由传统死区模型导致的控制奇异问题。同时,所提方法去除了控制器设计过程中已知物体尺寸参数的先验条件。应用李雅普诺夫稳定性理论,证明了所提方法在多机械臂协调操作过程中运动和内力控制性能。最后,仿真结果说明了所提方法的有效性。
[Abstract]:With the cross-disciplinary development of information, machinery, materials and so on, robots have shown good application potential and strong market demand in assisting or even replacing human beings in coordinating work. Therefore, the research of robot technology has not only obvious application prospects, but also important theoretical value. It is described as a multi-degree-of-freedom motion/force hybrid nonlinear system, focusing on how to deal with the uncertainties such as actuator nonlinearity and state constraints, aiming at improving the control performance of the robot. The structure of this paper is arranged as follows. Chapter 1 describes the relevant research background and research significance. Chapter 2 outlines the modeling and control of the robot system. Chapters 3 to 7 consist of five chapters, which are the main contents of this paper, including three aspects: 1) Chapters 3 to 4 mainly study the adaptive fuzzy coordinated control of robot with actuator nonlinearity; 2) Chapter 5 studies the adaptive control of generalized actuator nonlinearity robot based on Nussbaum function method. Chapters 6 to 7 mainly study the adaptive neural network coordinated control with state constraints. Specifically, these five chapters correspond to the following contents in turn: 1. The coordinated control problem of robot grasping object in the case of actuator clearance nonlinearity is studied, and a motion/force compensating actuator clearance is proposed. Firstly, based on the idea of backlash nonlinear inverse compensation, an inverse model adaptive control method for actuator backlash is constructed. Then, a decentralized robust adaptive fuzzy coordination control method is established to ensure that the motion and internal force of the object converge to the expected value respectively. Finally, the proposed method is applied to a two-arm robot system. The results of simulation and comparison with the existing methods show that the proposed method is effective. Secondly, the coordinated control problem of multi-manipulator under actuator hysteresis nonlinearity and motion constraints is studied, and a robot adaptive fuzzy control scheme based on Barrier Lyapunov function method is proposed. The hysteresis model is established in the dynamic equation of the manipulator, and then the adaptive control technique is introduced to compensate and reduce the influence of the unknown hysteresis nonlinearity. The qualitative theorem guarantees the motion and force control performance of the proposed method in the coordination process of multiple manipulators. Finally, several groups of comparison results show the effectiveness of the proposed method. The proposed method not only extends the unknown control coefficients from constants to time variables, but also removes the known assumptions of the upper and lower bounds of the control coefficients. The state of the robot system converges asymptotically to the desired trajectory in the case of generalized actuator nonlinearity. Furthermore, to reduce the control jitter caused by the use of traditional Nussbaum functions, a control method based on saturated Nussbaum functions is proposed. The proposed Nussbaum functions are constructed based on the idea of time expansion and reduce the traditional amplitude expansion Nus. In addition, by combining with the adaptive control method, a control method is established to deal with multiple unknown time-varying control coefficients, which facilitates the stability analysis of MIMO systems and guarantees the asymptotic tracking of the motion state of the robot system under unknown actuator dynamics. Fourthly, the problem of state hysteresis constraints caused by the output mechanism in the coordinated control of two manipulators is studied. An adaptive neural network controller is proposed to realize the coordinated control of robot motion and force. At the same time, combining with the adaptive neural network control method, the upper bound of neural network weight matrix is estimated, the number of adaptive laws to be updated is reduced, and the computational load to complete the real-time control is reduced. The results of performance comparison and evaluation further validate the effectiveness, superiority and robustness of the proposed method. Fifthly, the problem of coordinated control of multi-manipulators with unknown output dead-time constraints and uncertainties is studied, and a motion/force adaptive neural network coordinated control method for multi-manipulators is proposed. By using the Lyapunov stability theory, the motion and internal force control in the coordinated operation of multiple manipulators are proved. Finally, simulation results illustrate the effectiveness of the proposed method.
【学位授予单位】：广东工业大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TP242

【相似文献】