基于协作机器人仿真环境的关节轨迹预测方法

Joint Trajectory Prediction Method Based on Collaborative Robot Simulation Environment

安全在人机协作过程中是至关重要的,必须实时掌握人的行为信息,并进行准确高效的预测。基于Linux和ROS系统搭建仿真环境,通过Xtion PRO LIVE深度相机采集多组人体关节的空间位置信息,然后通过无监督学习方法对采集到的坐标点进行聚类和预测,实时更新预测模型,并基于minimum-jerk对特殊异常轨迹进行预测。为了充分保证人的安全,主要研究手部和肘部运动轨迹的预测方法。最终实验结果证明,所提出的分层轨迹预测框架可以很好地描述人体运动轨迹,并实时做出准确的预测,不仅保证了人体安全,而且对于提高生产效率具有重要意义。

Security is of vital importance in human robot collaboration,it is necessary to realize human behavior in real time and make accurate and efficient prediction.A simulation environment was built based on Linux and ROS system.The spatial position of multiple groups of human joints was collected by depth camera of Xtion PRO LIVE.Then,the unsupervised method was used to cluster and predict the collected coordinate points,and the prediction model was updated in real time,minimum-jerk was used to predict the special abnormal trajectory.To ensure the safety of human,the prediction methods of hand and elbow movement trajectory were studied.The final experimental results show that the hierarchical trajectory prediction framework proposed can describe the trajectory of human better and make accurate prediction in real time,which not only ensures the safety of human,but also plays an important role in improving production efficiency.