基于激光深度传感器的仿人机器人动作示教

针对机器人的运动示教问题，提出了一种基于激光深度传感器的示教系统。利用激光深度传感器能够快速准确获得人体各关节空间三维坐标的性质。通过分析机器人的关节构型，推导出由笛卡尔坐标空间到关节空间的转换公式，利用该公式，能够实现人体各关节转角的实时计算，进而实现对机器人的简单快捷的运动示教。为了在平滑数据的同时避免延时，提出了一种基于速度的有选择均值滤波法。将该示教方法应用在BioROBO仿人机器人上，试验结果证明了该方法的有效性。

Motion Teaching for Humanoid Robot Based on Laser Depth Sensor

A laser depth sensor based robot possesses the merits of getting human body joints＇ teaching system was proposed in this paper. The laser depth sensor 3D coordinates in real time accurately. Taking this advantage, and through the analysis of the robot＇s joint configuration, the transformation formula from the Cartesian coordinate space to the joint space was deduced. With this formula, the rotation angle of each joint was computed and the robot teaching was fulfilled. To smooth data without causing delay, a velocity-based selective mean filter （VSMF） was proposed. The teaching system was applied on BioROBO humanoid robots and the experiment results verified the effectiveness.