【作者】 丁度坤；

【导师】 谢存禧；

【作者基本信息】 华南理工大学 ， 机械设计及理论， 2010， 博士

【摘要】 工业机器人工作在结构化的工业现场环境中,可将人类从繁重、单调与重复的体力劳动中部分解脱出来,目前已大量应用于汽车、电子、信息产品等现代制造业。中国作为制造业大国,在走向制造业强国的过程中,工业机器人已成为其中不可或缺的重要组成部分。工业机器人是一个多输入多输出的非线性系统,具有时变、强耦合和非线性的特点。目前工业机器人控制器通常忽略了机器人的动力学效应,但是当工业机器人高速运动时,其动力学效应十分显著,不可忽略。这就要求控制器在的速度和有效载荷变化的情况下实时改变其增益系数,以提高机器人的控制精度。本论文分别对机器人工作轨迹已知情况下的高速运动以及工作轨迹未知条件下的高速运动控制问题开展研究。本课题研究项目获国家863计划重点项目(项目编号:2009AA043901)资助。论文针对负载为10kg的搬运机器人平台,进行了运动学、静力学与动力学分析。提出了多关节传递函数的推导方法。代入由轨迹而获得的部分数值解,对机器人的动力学方程进行简化,进而建立机器人的多关节传递函数表达式,将关节间的动力学效应包含入传递函数之中。本文提出了构建机器人关节非线性PID以及神经网络模型的新方法,解决“工作轨迹已知条件下”搬运机器人的高速运动控制,其中“工作轨迹已知”对于搬运机器人是指物料的取放位置确定,工作轨迹通过示教获得。对高速运动情况下的机器人动力学参量进行深入分析,得出影响机器人动力学效应的状态参量—机器人的关节角位置向量θ、角速度向量θ?与角加速度向量θ??。结合研究项目机器人样机的PID主控制器,探索机器人动力学状态参量与PID增益系数间的关系。首先通过实验,操纵机器人末端沿已知的轨迹运动,利用光电编码器在采样时刻内获取机器人在采样点的角位置、角速度与角加速度信息,并同时获取机器人在采样点的PID增益系数。然后初步选取了机器人的关节角位置向量θ作为影响PID增益变化的状态参量,利用最小二乘法推导了机器人的PID增益系数与各关节角位置的非线性表达式,构建机器人的关节非线性PID控制器。在此基础上,进一步考虑机器人关节角速度向量θ?与角加速度向量θ??对控制器参数的影响。利用神经网络建模技术构建了机器人的动力学状态参量(关节角位置向量θ、角速度向量θ?与角加速度向量θ?? )与PID增益系数(比例系数k p、积分系数k i与微分系数k d)间的BP神经网络模型。论文设计了改进型的免疫克隆与DNA控制器,解决“工作轨迹未知条件下”搬运机器人的高速运动控制。其中“工作轨迹未知”对于搬运机器人是指对散乱物料的高速抓取功能,工作轨迹需由相关传感器(如视觉传感器)实时检测获得。针对工作轨迹未知条件下,机器人的关节角位置、角速度与角加速度只能实时获取;要达到高速高精度目标则要求机器人的主控制器参数必须时刻最优的特点。提出机器人关节模型的实时辨识加上控制器增益系数实时自整定的控制方案。基于已建立的关节传递函数,利用加权指数最小二乘法实时辨识机器人关节的被控对象模型。进而借鉴了DNA计算与免疫算法工作机制,并对其进行了一些局部的改进,创造性地设计了改进型的免疫克隆与DNA控制器,用于控制器增益系数的实时自整定。这种控制方案能较好地抑制工作现场各种不确定性干扰的影响,同时又能确保控制器参数的时刻最优性,且具有一定的自适应性与智能性,适用于运动轨迹时变、由机器人本身去实时识别轨迹的高速工作场合。论文进行了相关的仿真与运动控制对比实验,利用Visual C#高级编程语言实现了所研究的控制算法,将其嵌入控制系统中,进行了相关的实验分析。实验结果表明,新方法较传统的单关节PID控制在速度、精度方面均有较大提高。对于非线性PID控制与神经网络模型,其算法运算时间分别为1.8ms与3ms,跟踪精度分别为±0.28mm与±0.1mm,神经网络模型比非线性PID的跟踪精度有较大提高,实时性好,适用于工作轨迹固定、通过示教获知轨迹的机器人高速运动控制。对于改进后的免疫克隆与DNA计算,其算法运算时间为72ms与50ms,跟踪精度分别为±0.08mm与±0.06mm,改进后的DNA计算在算法的实时性与跟踪精度方面均较免疫算法有提高,适用于工作轨迹多变、由机器人自己实时识别轨迹的高速运动控制。本论文的创造性成果有:1、机器人关节建模方面,不同于传统的单关节建模方式,考虑了机器人关节间的耦合效应,推导机器人多关节传递函数模型。2、对机器人的动力学表达式进行了深入分析,得出了影响机器人动力学效应的状态参量—机器人的关节角位置向量θ、角速度向量θ?与角加速度向量θ??。3、结合项目机器人样机的主PID控制器,定性与定量探索机器人的动力学状态参量与控制器增益之间的关系规律,设计了非线性PID控制器与神经网络模型。4、设计了改进型的免疫克隆与DNA计算算法。论文的研究成果可以解决机器人高速高精度运动难题,促进机械工业中自动化的发展和社会生产率的提高更多还原

【Abstract】 Industry robot works in a structural environment and can free human being from the heavy, boring, repetition physical labor. Nowadays, robot has been widely used in the field of automobile, electronics, information production and so on. China as a big production country, industry robot has become its important part during the process to be a strong production country. Industry robot is a very complicated non-linear system, which has multiple input and output. It has the character of time-varying, strong couple and non-linear. The controller, which is used in robot nowadays, usually ignores the robot’s dynamic effect. However, when robot runs in the high speed condition, its dynamics effect is very huge, which can’t be ignored. At this time, the speed and payload of robot varies greatly. In order to improve robot’s working accuracy, the controller’s parameters, should change, too. Therefore, to solve the control problems when robot runs at a high speed under an unknown working trajectory and runs at a high speed under a known working trajectory, the paper make a deeply research. The research work has been supported by the national 863 important project. The project number is 2009AA043901.To the transport robot whose payload is 10kg, the kinematics, static and dynamics analysis of the robot are performed in this paper. Different from the traditional single joint modeling, the paper proposed a new method to set up the multiple joints’transfer function. The dynamic function of robot can be simplified by some numerical solution obtained in the trajectory and the multiple joints’transfer functions can be set up, which takes the joints’dynamic effect into consideration.The paper proposed a new method to setup a joints’nonlinear PID controller and neural network, to solve the problem when robot run in a high speed condition whose working trajectory is known. To transport robot, the working trajectory is known means that the material’s putting position is determined and can be obtained by teaching. The dynamic parameters of robot have been deeply researched when it runs under a high speed condition. Then the state parameters affecting the robot’s dynamic effect can be obtained, which is the robot’s angular position vetorθ, the angular speed vectorθ? and the angular acceleration vectorθ?? .To the PID main controller of the robot platform in the project, the relationship between the dynamic parameters and the PID coefficients has been deeply researched.The tip of the robot is firstly manipulated running along a known trajectory in experiment. At the sample time, the robot’s joints’angular position, the angular velocity and the angular acceleration can be obtained by the photoelectric encoder. The PID parameters at the sampling point can be attained, too. The joints’angular positionθis firstly chosen to be the controller’s state parameter. Then the nonlinear expression between the robot’s PID parameters and the joints’angular position can be inferred by the least square method. So the robot’s joints’nonlinear PID controllers can be set up.Moreover, the robot joints’angular speed vectorθ? and the angular acceleration vectorθ?? , which may affect the PID controllers, have been taken into consideration. The neural network intelligent modeling technology is used to setup the BP neural model between the robot’s joints’angular position vectorθ, angular speed vectorθ?, angular accelerationθ?? the PID parameters k p, k i, k d.The paper has designed the improving immune clone select and DNA controllers to deal with the problem when the transport robot working in a high speed motion condition whose trajectory is unknown. To the transport robot, the working trajectory is unknown means that robot fetches the scattering material in high speed and the trajectory can be detected in real-time by some sensors, such as the visual sensor and so on. Due to the unknown working trajectory means that the robot joints’angular position, angular velocity and angular acceleration can only obtained in real time. The high accuracy and high speed goal request the main controller’s parameters should optimistic at any time. Due to this characteristic, the paper proposed a control scheme which is the robot joints’model’s real-time recognition and the controller’s parameters’real-time self tuning.Based on the joints’transfer functions set up, the model of the robot’s joints can be identified by the weighted type least square method. The working mechanism of the DNA compute and immune algorithm has been deeply researched and improve partly. Then the improving immune clone select and improving DNA algorithm have been creatively designed and used to tune the controllers’parameters in real time. The control scheme can restrain the effect of various uncertain factors’disturb and can make sure the control parameter’s optimality at any time, which has some adaptability and intelligence. This control scheme can be used in the condition that the moving trajectory is time-varying which is recognized by the robot itself in real time, and the demand of accuracy and speed is very high.In the end of the paper, several simulation experiments have been caught out to test the researched intelligent control algorithm. All of the algorithm are coded by the Visual C# program language and embedded into the control system. Then some motion control experiments are performed. The final experiment results show that, the new method is better than the traditional single PID control in the aspect of speed and accuracy. The running time of nonlinear PID control and the neural network model is 1.8ms and 3ms. The final tracking error of the algorithm is±0.28mm and±0.1mm. The neural network model has a stronger real-timing character, which can be used to several robots requiring running in a high speed condition when working trajectory is regular and can be known by teaching.The running time of improving Immune clone selection algorithm and DNA compute is 72ms and 50ms, and the final tracking error is±0.08mm and±0.06mm. The improving DNA compute need less running time and can reached a higher tracking accuracy than that of the immune clone select algorithm, which is suitable for high speed motion command when working trajectory is varying and be recognized by the robot itselfThe creative achievements can be shown as following:1. In the aspect of robot joints’modeling, different from the traditional modeling manner, the robot’s multiple joints transfer functions have been inferred, which take the robot joints’coupling effect into consideration.2. The dynamic expression has been deeply analyzed, and the state parameters affecting the robot’s dynamic effect are obtained, which are the robot’s joint angular position vectorθ, the angular velocity vectorθ? and the angular acceleration vectorθ?? .3. To the robot platform’s main PID controllers, the relationship between the robot dynamic state parameters and the controllers’coefficients in qualitative and quantitative analysis.4. The improving immune select clone algorithm and DNA computing algorithm.The research achievement can solve the problem of the high accuracy requirement when robot run in a high speed motion condition, promote the automation in mechanic industry and boost the social productivity.更多还原