Abstract: To ensure the efficient and safe process of human-robot collaboration tasks,a safe human-robot collaboration method based on abnormal behavior monitoring was proposed.The human motion feature model was constructed based on 3D skeleton features of human body,and the motion feature data were de-anomaly and filtered.The standard operating behavior was learned iteratively from the human demonstration and was combined with the minimum human-robot distance to identify and classify the abnormal behavior online.Based on the artificial potential field algorithm,a collision avoidance motion planning was performed for the robot under the abnormal behavior to prevent dangerous human-robot collision.The proposed method was validated by part sorting and human-robot collaborative gluing tasks,and the experimental results showed that the method could effectively ensure the safety of the human-robot collaborative process by accurately monitoring abnormal behaviors.

Key words: human-robot collaboration, minimum distance, motion planning, artificial potential field, collision avoidance

摘要： 为保证人机协作任务的高效安全进行,提出一种基于异常行为监测的人机安全协作方法。首先,根据人体3D骨架特征,构建人体运动特征模型,并对运动特征数据进行去异常点和滤波处理;其次,从人员示范中迭代学习标准作业行为,结合人机最小距离,对异常行为进行在线识别与分类;最后,基于人工势场法,在异常行为下对机器人进行避碰运动规划,防止人机危险碰撞。通过零件分拣和人机协作涂胶任务验证了所述方法,实验结果表明该方法可通过准确监测异常行为,有效保证人机协作过程安全。

关键词: 人机协作, 最小距离, 运动规划, 人工势场法, 避碰