Tactile Sensing Analysis Of Robot Hand Based On Mach-zehnder Interferometer

With the widespread application of robots in various fields,the functional requirements of robot tactile sensing are also increasing.Compared with the traditional electrical tactile sensors,distributed optical fiber sensors have attracted wide attention in various industries due to their advantages of continuous sensing,high flexibility and low electromagnetic interference.Among them,Mach-Zehnder interferometer has the advantages of simple structure,high sensitivity and high real-time performance,and it has great application prospect in robot tactile sensing.In view of the limitations of the existing optical fiber tactile sensors that the structures are complex and easy to be interfered by light intensity,this paper builds a simple structure and high sensitivity Mach-Zehnder optical fiber sensing system to verify the tactile sensing function of the bionic robot hand.In addition,in order to restore external disturbance signal accurately,we adopt an improved the arctangent demodulation method for signal demodulation.Finally,a signal identification model based on empirical mode decomposition method and BP neural network is established to classify and identify the demodulated signals.The advantages and feasibility of the tactile sensing system proposed in this paper are verified by simulations and experiments.Firstly,this paper introduces the research background of robotic hand tactile sensing,the research status of optical fiber tactile sensors,the development of distributed optical fiber sensing and demodulation technologies and signal recognition methods,and points out the advantages of Mach-Zehnder optical fiber tactile sensing system proposed in this paper.Then,the basic theories of Mach-Zehnder optical fiber sensing system are described in detail.In order to solve the problems that the traditional arctangent demodulation method is prone to produce harmonic distortion and phase winding,this paper improves the method by differential self-multiplication and phase unwrapping algorithm respectively.In addition,a BP neural network prediction model is established for signal continuation to eliminate the influence of the end effect of empirical mode decomposition on the decomposition results.Finally,the improved arctangent demodulation method and identification model proposed in this paper are simulated and experimented on the MATLAB platform.The simulation results show that the improved arctangent demodulation method and identification model can effectively improve the accuracies of signal demodulation and identification.The experimental results show that the average recognition accuracy of the signal recognition scheme proposed in this paper is as high as 97.87%.