面向直接示教的机器人负载自适应零力控制

Load Adaptive Force-free Control for the Direct Teaching of Robots

  • 摘要: 相对于基于末端多维力传感器的直接示教工业机器人,基于关节扭矩传感器信息的机器人可接触范围大,且安全性更好.但在变负载的复杂工况下,加上受到机器人自身摩擦力、重力、惯性力的影响,示教效率受到制约.为此,本文提出了一种零力控制方法.首先,引入柔性关节机械臂的动力学模型作为被控对象,并分析了该模型中电机摩擦力、惯量以及机械臂重力在直接示教过程中的影响.然后,为精确补偿机器人自身的重力,基于QR分解与最小二乘理论进行参数辨识.而且针对机器人更换末端执行器或抓取物品后模型参数发生变化的情况,提出一种变负载自适应零力控制方法.最后,在自主研发的7自由度协作型机器人平台上进行了实验.机械臂自身重力参数辨识后的模型最大计算误差在关节额定力矩的4%以内,单关节力牵引实验中通过调节零力控制参数,牵引力矩可由约13 N·m下降到约2 N·m.变负载下的机器人牵引实验中,控制器能够在10s内更新参数.实验结果表明,在负载变化的情况下,该控制方法可帮助操作者轻松地拖拽机器人进行精准、高效的直接示教.

     

    Abstract: Comparing with the direct teaching robots with a multi-dimensional force sensor mounted on the end-effector, the robots based on the information from the joint torque sensors, have wider accessible area and are superior in security. But when facing complex tasks with variable loads, the teaching efficiency is constrained, and also affected by the friction force, the gravity and the inertia force on the robot itself.Therefore, a force-free control method is proposed. Firstly, the dynamic model of the flexible-joint manipulator is introduced as the controlled object. The influences of the friction force and the inertia of the motor are analyzed, as well as the gravity of the manipulator in the process of direct teaching. Then, its parameters are identified using QR decomposition and least squares theory to accurately compensate the gravity of the robot. Furthermore, a variable-load adaptive force-free control method is proposed to solve the model parameter change problem caused by replacing the end effector or grasping an object. Finally, experiments are carried out on the self-developed 7-DOF cooperative robot platform. The maximum calculation error of the model is less than 4% of the joint rated torque, after identifying the gravity parameter of the robot itself. In the experiment of single joint force traction, the moment is reduced from about 13 N·m to about 2 N·m by adjusting the force-free control parameters. In robot traction experiment with the variable load, the controller can update the model parameters within 10s. The experiment results show that the control method can help operators to drag the robot easily for an accurate and efficient direct teaching in the case of the load changes.

     

/

返回文章
返回