【作者】 高鹏；

【导师】 张平；

【作者基本信息】 华南理工大学 ， 计算机应用技术， 2015， 硕士

【摘要】 随着机器人技术在各个领域的广泛应用,机器人与人类的接触越来越不可避免,为保障人类的安全,同时提高机器人的智能性,开展人机协作技术及其应用研究逐渐成为研究热点。传统的机器人避碰研究主要是对物体的避碰,本文提出了一种人机协作场景下的机器人主动避碰的方法,使机器人智能地识别并躲避人体。本文的主要研究内容如下:1)基于多Kinect传感器提取系统获得人体与机器人的三维结构信息以减小各种障碍遮挡对人体与机器人的三维结构数据采集造成的影响;根据人体特征点信息构建人体特征区域,建立了人体仿真模型。2)基于卡尔曼滤波对人体骨架位置进行预测,并结合人体结构特征定义人体关节间的约束,提高对人体三维信息的数据预测精度。3)基于roadmap思想,在线性空间内解决机器人路径规划的问题;采用分层碰撞检测技术进行人-机器人碰撞检测,降低机器人实时避碰过程中的计算代价的同时保证人机安全。4)基于决策表思想,设计机器人主动避碰规则,从人体位置以及运动速度等方面决策机器人的避碰动作,提高了机器人主动避碰的智能性,减少机器人无谓的运动,保证避碰效率。本文最后设计机器人主动避碰实验对上述内容进行仿真和实验验证,实验结果表明本文提出的主动避碰方法有效、合理。更多还原

【Abstract】 With the extensive application of robots,the contact of robots and human is more andmore inevitable.To ensure security of human while increasing the intelligence ofrobots,carrying out human-machine collaboration technologies and its applied research hasgradually become a hot topic.Traditional research on collision avoidance of robot mainly focuses on static object,thispaper gives a way in the man-machine collaboration scene,aiming at making robot recognizehuman and avoid collision with human..The main work of this paper is as follows:1)Extractthe three-dimensional information of human and robot base on Kinect sensor which has goodinformation acquisition ability, reduce the impact of the data coherence caused by occlusionusing multiple Kinect sensors,construct human characteristics region base on the featurepoints and make simulation of human body model.2)Make prediction of position of humanskeletons base on the Kalman filter,define constraints between human joints using unscentedKalman filter,combined with the physiological characteristics,to improve the dataaccuracy.3)Based on the roadmap idea,this paper solves the problem of robot path planning ina flat space,reduce the computational cost of the robot in the process of real-time collisionavoidance.This paper uses hierarchical collision detection technology to ensure safety ofhuman-machine.4)Base on the decision table idea,this paper designs active collisionavoidance regulations of robots with decision-making of both speed and direction,improvingthe intelligence of robot initiative collision avoidance,reducing unnecessary movement of therobot and ensuring efficiency.Finally, the experiment of the robot collision avoidance isdesigned to make simulation and verification for the above.Experiment results show that theactive collision avoidance method proposed in this paper has validity and rationality.更多还原