摘要
针对人机协作时机器人极有可能与人发生碰撞所导致的安全问题,提出一种基于动力学模型、无需要外部传感器的机器人碰撞检测方案.首先,根据对机器人低速下的动力学特性分析,采用逐次动力学辨识,在对机器人重力及摩擦力使用最小二乘法辨识后,运用Lasso对复杂的残余力矩进行辨识以获得简化模型,避免了复杂符号推导.然后,针对模型的不确定性,根据优化设计运动状态相关的动态阈值来处理残差的异常峰值.最后,以轻型工业机械臂为实验平台,让机器人执行作业轨迹与人类发生碰撞以验证碰撞检测算法.实验结果证明了碰撞检测方法的可行性和简单性,其能够实现对机器人快速动力学辨识,并可以实现灵敏准确碰撞检测,保证人机协作时操作人员的安全.
Robot is more likely to collide with humans when HRC(human-robot collaboration)is demanded.To handle the safety issue of HRC,a collision detection method is proposed based on dynamic model without employing exteroceptive sensors.Firstly,a sequential identification procedure is introduced based on the analysis of the dynamic property of robot with low velocity.After the torque effects of gravity and friction are estimated using least squares method,the remaining torque is predicted based on Lasso,which allows a simplified model expression and relief of complicated symbol computation.Subsequently,a motion state-dependant dynamic threshold is designed to tackle the abnormal peaks induced by the model uncertainties.Finally,collision between human and robot is imposed to verify the collision detection algorithm on a light industrial manipulator executing a task trajectory.The experimental results demonstrate the feasibility and simplicity of the proposed algorithm which enables fast dynamics identification and realizes sensitive and accurate collision recogition to guarantee operator’s safety for HRC scenario.
作者
甘亚辉
曹鹏飞
戴先中
GAN Ya-hui;CAO Peng-fei;DAI Xian-zhong(School of Automation,Key Laboratory of Measurement and Control of Complex Systems of Engineering,Southeast University,Nanjing Jiangsu 210096,China)
出处
《控制理论与应用》
EI
CAS
CSCD
北大核心
2019年第9期1509-1519,共11页
Control Theory & Applications
基金
国家自然科学基金项目(61873308,61503076,61175113)
江苏省自然科学基金项目(BK20150624)
复杂工程系统测量与控制教育部重点实验室开放课题(MCCSE2014B02)资助~~
关键词
无传感器碰撞检测
人机交互
动力学辨识
Lasso
sensorless collision detection
human-robot interaction
dynamics identification
Lasso
