基于内嵌型光学弯曲传感器的软体手感知

Soft hand perception based on embedded optical fiber bending sensor

软体仿人手的感知能力是实现友好交互和灵巧操作的重要条件.柔性的光学弯曲传感器由发光二极管(light emitting diode,LED)、光导纤维和光敏传感器组成,易嵌于软体手指内.经粗糙化的光导纤维可产生光线溢出,通过测量光能损耗量可获得软体手指的弯曲度;利用归一化标定软体手指初始(伸直)和最大弯曲状态的采样电压值,对归一化后的电压值进行均匀、非均匀和滞环分段,并融合5指弯曲信息实现连续运动检测、动作姿态和物体大小识别.实验结果表明,光学弯曲传感器可检测连续正弦运动状态;正确识别多种手势和抓握姿态;也可识别物体大小,结合滞环分段方法,解决单点分类切换存在振荡不足的问题,提高识别率.这种光学式柔性弯曲传感器具有电磁免疫、易于内嵌的优势,在软体手和软体执行器中具有重要应用价值.

The perception capability of soft humanoid hand is very important to the implementation of friendly interactive and dexterous operation. The flexible optical bending sensor which is easy to be embedded into the soft finger is comprised of light emitting diode( LED),optical fiber and photo-sensitive sensor. The roughened optical fibers can generate light overflow. And the bend degree of soft fingers can be obtained by detecting the loss of light energy. The abilities of continuous movement detection,action gestures and object size recognition are implemented according to the following operations:① calibrating the initial( unbent) and maximum bending statuses of soft fingers with normalization method;② segmenting the normalized data by uniform,nonuniform and hysteresis segment method;③ merging together the information of five fingers. The experiment results show that the proposed optical fiber bending sensor is able to detect the continuous sinusoidal movement status,correctly recognize various gestures and grasping postures and the object size. Combining hysteresis subsection method,the shortcoming of single point subsection oscillating near subsection point is overcome,and the recognition rate is improved. This optical flexible bending sensor has the advantages of electromagnetic immunity and easy to be embedded. It has important application value in soft robotic hand or soft actuator.