基于体感的机器人仿人动作模仿系统研究

发布时间：2018-05-09 22:01

本文选题：人机交互 + 人体动作模仿　；参考：《西华大学》2016年硕士论文

【摘要】：作为衡量现代科技水平的重要标志,机器人技术受到国内外越来越多的重视。目前,机器人的研究已经在运动能力、人机交互、人工智能等方面取得了较大的进步,但仍面临着许多技术难题。人体动作模仿可以提高机器人的人机交互能力、学习能力和适应能力,让机器人更好地服务于人类。国内外现有的人体动作模仿的研究在动作获取方式、运动的机动性以及动作模仿效果上都还存在局限。因此,本文利用体感技术来获取人体动作信息,并将其应用于仿人机器人和工业机器人的仿人运动控制。首先,利用位姿描述、D-H机器人坐标系建立法以及正运动学方程建立了仿人机器人NAO机器人右腿运动学模型,且分别使用解析法和基于LM算法的BP神经网络的方法来求解其逆运动学。利用微软体感传感器Kinect、NAO以及PC机,构建了仿人机器人姿态动作模仿系统,根据Kinect人体骨骼信息和NAO机器人关节模型的各自特点,设计并实现了一种基于解析法的人体姿态到NAO仿人机器人姿态的转化方法,从而实现机器人对人体姿态的模仿。然后,依据COM投影与支撑多边形的位置关系来判定机器人姿态的稳定性,提出了一种姿态稳定性修正策略,对不稳定姿态的脚掌及脚踝目标位置进行调整,并使用LM算法来求取姿态修正过程中的运动学逆解。再结合姿态序列滤波限速等步骤,设计并用程序实现了仿人机器人动作模仿控制流程。还分别使用关节角度和连杆方向两种方法,对模仿过程中机器人和人体的姿态相似性进行定量评价。实验结果表明,通过本文的姿态转化算法和动作模仿控制,机器人能够模仿出与人体相似的姿态和动作,且能保持运动过程中的稳定性。同时,本文的相似性测度进一步验证了动作模仿算法的有效性。此外,设计并实现了一个基于人体手势控制的SCARA机器人写字系统,利用Kinect获取人体姿态信息,并使用手掌的相对位置来控制机器人的绘制轨迹,通过手掌状态来控制机器人的提笔和落笔动作。为提高机器人模仿写字的控制效率,还提出了一种基于设定误差范围来处理人体绘制轨迹方法。实验表明,该系统操作方便,能够使机器人快速精确地模仿人的写字动作。
[Abstract]:As an important symbol to measure the level of modern science and technology, robot technology has been paid more and more attention at home and abroad. At present, the research of robot has made great progress in motion ability, human-computer interaction, artificial intelligence and so on, but it still faces many technical problems. Human body action imitation can improve the human-computer interaction ability, learning ability and adaptive ability of robot, so that robot can better serve human beings. At home and abroad, there are still limitations in the way of motion acquisition, the mobility of movement and the effect of motion imitation. Therefore, this paper uses the technology of body sensation to obtain human body action information, and applies it to the humanoid motion control of humanoid robot and industrial robot. Firstly, the kinematics model of the right leg of the humanoid robot NAO robot is established by using the position and pose description method of D-H robot coordinate system and the forward kinematics equation. The inverse kinematics is solved by analytic method and BP neural network method based on LM algorithm. A humanoid robot posture simulation system is constructed by using Microsoft Kinectnao and PC. According to the characteristics of Kinect human skeleton information and NAO robot joint model, the simulation system is designed. A transformation method from human attitude to NAO humanoid robot based on analytic method is designed and implemented, so that the robot can imitate human posture. Then, according to the position relationship between COM projection and support polygon, the stability of robot attitude is determined, and an attitude stability correction strategy is proposed, which adjusts the position of feet and ankles of unstable attitude. The inverse kinematics solution of attitude correction is obtained by using LM algorithm. Combined with the steps of attitude sequence filtering and speed limiting, the simulation control flow of humanoid robot is designed and implemented. The attitude similarity between robot and human is evaluated quantitatively by using joint angle and linkage direction respectively. The experimental results show that the robot can imitate the posture and movement similar to the human body and maintain the stability of the motion process through the attitude transformation algorithm and motion simulation control in this paper. At the same time, the similarity measure of this paper further verifies the effectiveness of the action imitation algorithm. In addition, a SCARA robot writing system based on human gesture control is designed and implemented. The posture information of human body is acquired by Kinect, and the relative position of the palm is used to control the robot's drawing trajectory. Control the lifting and writing movements of the robot through the palm of the hand. In order to improve the control efficiency of robot imitating writing, a method to deal with the human body drawing trajectory based on setting error range is proposed. The experiment shows that the system is easy to operate and can make the robot imitate the action of human writing quickly and accurately.
【学位授予单位】：西华大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP242

【相似文献】