人类行为模式启发的人机物体传递

A Human-to-robot Handover Method Inspired by Human Behavior Patterns

针对目前人到机器人的物体传递过程中,机器人难以像人类一样理解给予者的意图并配合完成传递任务的问题,提出了一种人类行为模式启发的人机物体传递方法。通过观察人类物体传递过程总结了人类的物体传递行为模式,使用骨骼信息与RGB-D信息融合构建传递意图判别模型;控制机器人在识别到传递意图后迅速配合完成传递任务;通过相关实验验证了该方法在人机传递效率上的改善,并基于机器人社会属性量表(RoSAS)面向使用者开展了问卷调查。实验表明,文章提出的传递意图识别准确率为96.2%,且有效的区分了相同姿态表达的不同意图,人机传递系统成功率为91.1%,平均传递时间3.3 s,能够满足人机物体传递中准确性与流畅性的需求。

In view of the current robot′s inability to understand the giver′s potential intention as well as the human being in the handover process,this paper proposed a human-to-robot handover method inspired by human behavior patterns.This paper first summarized the human handover behavior pattern by observing the relevant process,and constructed the delivery intention discrimination model by using the skeletion information and RGB-D information.Then the robot was controlled to complete the handover task after recognizing the delivery intention.Finally,the improvement of human-to-robot handover efficiency was verified by relevant experiments,and a questionnaire survey based on the Robot Social Attribute Scale(RoSAS)was conducted among the users.The experiment shows that the accuracy of the delivery intent recognition proposed in this paper is 96.2%,and it effectively distinguishes different intentions expressed in the same pose.The success rate of the human-to-robot handover system is 91.1%,and the average handover time is 3.3 s,which can meet the needs of accuracy and fluency in human-to-robot handover.