机器人力-位置智能控制关键技术与应用的研究

项目名称： 机器人力-位置智能控制关键技术与应用的研究

项目编号： No.61473200

项目类型： 面上项目

立项/批准年度： 2015

项目学科： 自动化技术、计算机技术

项目作者： 谢小辉

作者单位： 苏州大学

项目金额： 82万元

中文摘要： 针对一类工业机器人打磨系统阻抗控制器中的实际检测接触力很难直接获得,本项目采用采集各伺服电机转矩值经过换算得到末端执行器与外部环境的接触力的方法,来替代直接采用力传感器信号反馈末端执行器的接触力。采用此类方法，可降低成本、提高效率、增加可靠性；但存在力传递及解算过程中的反向间隙、较大的摩擦力干扰、迟滞等问题。为解决上述问题，本项目在无源性理论的基础上,提出采用基于关节转矩解算方程的神经网络预测器预测反馈力, 并通过能量均衡器对预测力值和当前解算力值进行能量校正,使打磨机器人控制系统保持无源性且具备主动柔顺的能力。对期望接触力,则设计一种实验方法得到打磨力、打磨时间和在两种因素作用下的被打磨厚度,预估出期望的打磨力范围,据此对阻抗控制器的位置环输出在线进行局部的微量调整,以满足打磨要求。在自主设计的机器人抛光打磨系统中，通过锌合金水龙头把手的打磨实验来证明设计方法的可行性。

中文关键词： 机器人控制；主动柔顺；无源性；神经网络；打磨

英文摘要： For a class of industrial robot grinding system, the actual value of contact force is difficult to obtain by the system impedance controller. This project adopts method of gathering the servo motor torque value to compute the contact force between the end effector and the external environment, instead of directly using sensor to detect the contact force of the end effector. Using such methods, we can reduce cost, improve efficiency and increase reliability, but there are some problem like the reverse gap, high friction interference and hysteresis in the process of force transmission and computation. In order to solve such problem, using passivity theory this project gives a neural network predictor based on joint torque computing equations to predict the feedback force, and use the energy correction module to balance the predict value and the current solution value to achive energy calibration. This method makes the control system of the polishing robot remain passive and get the ability of active compliance. This project designs an experimental method to get the relation between the two factors of grinding force, grinding time and the thickness of grinding. By such experiments, we can estimate the expected range of the grinding force. According to the grinding data, the output of position loop of the impedance controller can be micro adjustmented online to meet the requirements of actual grinding. The feasibility of presented mthods will be proved by some grinding experiments of zinc alloy faucet handle on our designed robot grinding system.

英文关键词： Robot control;Active compliance;Passivity;Neural network;Grinding