College of Instrumentation and Electrical Engineering, Jilin University, Changchun 130061, China.
Jilin Insitute of Metrology, Changchun 130103, China.
Rev Sci Instrum. 2021 Sep 1;92(9):095003. doi: 10.1063/5.0057236.
Based on the tactile mechanism of human fingertips, a bionic tactile sensor fabricated from polyvinylidene fluoride piezoelectric film is proposed, which can identify the surface softness, viscoelasticity, thermal conductivity, and texture roughness of the object. The tactile sensor is mounted on the fingertip of the bionic manipulator, which obtains the surface features by touching and sliding the object. The time-domain features of the output signal are used for preliminarily discriminating the softness, viscoelasticity, and heat conduction of the object. Finally, based on the Back Propagation and the Particle Swarm Optimization-Back Propagation neural network algorithm, the recognition experiment of texture roughness is carried out using the PSO algorithm to improve the BP neural network so that the optimized BP algorithm has a higher convergence accuracy. The results show that the PSO-BP algorithm achieved the highest accuracy of 98% for identifying samples with different roughnesses and the average recognition achieved an accuracy of 94%. The bionic piezoelectric tactile sensor proposed in this paper has a good application development prospect in recognizing the surface features of objects and intelligent robots.
基于人类指尖的触觉机理，提出了一种由聚偏氟乙烯压电薄膜制成的仿生触觉传感器，该传感器可以识别物体的表面柔软度、粘弹性、导热性和纹理粗糙度。触觉传感器安装在仿生机械手的指尖上，通过触摸和滑动物体来获取表面特征。输出信号的时域特征用于初步区分物体的柔软度、粘弹性和导热性。最后，基于反向传播和粒子群优化-反向传播神经网络算法，使用 PSO 算法对纹理粗糙度的识别实验进行了优化，从而提高了 BP 神经网络的优化 BP 算法的收敛精度更高。结果表明，PSO-BP 算法对不同粗糙度的样本的识别精度最高，达到 98%，平均识别精度达到 94%。本文提出的仿生压电触觉传感器在识别物体表面特征和智能机器人方面具有良好的应用开发前景。
